First Digit Probability and Benford's Law

Probability Surveys
Vol. 0 (0000) 1–0

ISSN: 1549-5787
DOI: 10.1214/154957804100000000
First Digit Probability and Benford’s

Law∗
Irina Pashchenko
e-mail: [email protected]
Abstract: The following work shows how the first digit probability of a
group of numbers in certain situations that are close to real-life ones can
be explained using basic continuous real-valued functions. For instance,
the first digits probability of the numbers representing the change in human
growth can be understood better by looking at the square root function in a
particular way. Also, examples explaining the first digit probability of other
real-valued functions and how they are related to specific real-life situations
were brought to the reader’s attention. Moreover, a comprehensive analysis
of a group of certain discrete functions was performed by approximating
the functions to the above mentioned continuous ones, taking limits, and
other methods.
MSC 2010 subject classifications: Primary 60-01; secondary 62E99.
Keywords and phrases: Benford’s Law, first digit probability.
Received January 0000.
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Fdp of basic continuous functions . . . . . . . . . . . . . . . . 7
2.1 Pk of the exponential function y = ax . . . . . . . . . . . 7
2.2 Pk of the power function y = xa . . . . . . . . . . . . . . 17
2.3 Pk of the linear function y =√mx . . . . . . . . . . . . . . 26
2.4 Pk of the root function y = a x . . . . . . . . . . . . . . . 30
2.5 Pk of the logarithmic function y = loga x . . . . . . . . . 41
2.6 Pk of the reciprocal function y = xa . . . . . . . . . . . . 53
3 First digit probability of basic discrete functions . . . . . . . 63
3.1 First digit probability of prime numbers . . . . . . . . 63
3.2 First digit probability of the Fibonacci function . . . . 67
3.3 First digit probability of the factorial function . . . . . 70
4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . 81
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
∗ This is an original survey paper
1
I. Pashchenko/First Digit Probability and Benford’s Law 2
1. Introduction
Since our goal is to see how a group of functions can describe real-life situations,
all the functions will be graphed and analyzed in the first quadrant only since
this quadrant provides us with positive x- and y-values.
Also, let us acknowledge that if we are talking about the first digit of a
number, we mean the first digit of the whole portion of a number in a case if
the number is a decimal. In other words, we assume that the number is bigger
or equal to the number 1. That is why the range of all the functions that will
be discussed later should be restricted to [1, ∞).
The domain of each function will be restricted to match the corresponding
restricted range. For instance, the domain of the exponential function will be
restricted to [0, ∞) to match the range mentioned above.
If a random real number is selected in the [1, ∞) interval, the chance of getting
a number which starts with a particular digit will be equal to the chance of
getting a number which starts with any other digit. However, the vast majority
of the functions that will be discussed in this work will have a different outcome.
We will see how the probability of each first digit depends on the functions shape.
In order to be able to calculate the first digit probability of each functions
y-value, a particular ruler is proposed. The ruler takes the interval [1, ∞) and
has infinitely many sections in it. Each section is exactly ten times longer than
the previous one.
Moreover, each section is divided into nine equal parts. Each part is colored
with a particular color. Connecting all the parts and sections together brings
us back to the interval [1, ∞) without any overlapping or gaps. The same color
is used to label all the intervals containing numbers that start with the same
digit. Therefore, nine different colors were used for the infinitely long ruler.
The limited part of the ruler shown in Figure 1 can be used to calculate the
first digit probability in an interval [1, 100). Each color will represent a total
group of y-values that start with a particular digit. The complete ruler can be
applied for any real-valued function on a specified range [1, ∞).
The first section of the ruler is shown in Figure 2. It starts at 1 and ends at 10.
In order to calculate the first digits probability fairly, the number 1 is included
to the interval, but the number 10 is not. Therefore, the interval is [1, 10). A
randomly chosen number from this interval will have the same chance of having
any of the nine digits as the leading ones because the interval is broken into
nine pieces of equal length. Each of them starts at a closed point representing
a whole number from 1 to 9 and ends at an open point representing the next
whole number from 2 to 10. This fact makes all the chances of picking a number,
which start with a particular digit equal to each other.
Fig 1. The measuring ruler for the range [1, 100).

The same partition will be used for the rest of the sections of the ruler. The
interval [10, 100) is shown below in Figure 3. It is also broken into nine equal
intervals in the same manner. As it was mentioned above, the section [10, 100)
is exactly ten times longer than the previous one.
Any of the sections of the ruler can be expressed using a general form. It is
[10n−1 , 10n ), where n is a natural number. In particular, all the parts of the
section can be listed as the following intervals:
[1 · 10n−1 , 2 · 10n−1 ), [2 · 10n−1 , 3 · 10n−1 ), [3 · 10n−1 , 4 · 10n−1 ), ..., [9 · 10n−1 , 10n ).
Again, the chances of picking a number with a particular first digit from the
interval [10n−1 , 10n ) are equal to each other due to the equal lengths of the
subintervals. The interval is shown in Figure 4.
Fig 4. The measuring ruler for the range [10n−1 , 10n ).
In order to compare each function’s first digit probability to each other, a term
of the first digit probability will be used. The abbreviation f dp for discussion
purposes and Pk for formulas where k is the first digit of a number will be
utilized in this work. Let us recall that when any event takes place (rolling a
die, etc.), the probability of each single outcome of the event is a fraction from 0
to 1 which may be equal to either 0 or 1 as well and the sum of the probabilities
of all the outcomes equals to 1. Since any number of any real-valued functions
range starts with a digit from 1 to 9, the probability of any of the digits to
be the first one is located in an interval 0 ≤ Pk ≤ 1 where k = 1, 2, , 9 and
P1 + P2 + P3 + P4 + P5 + P6 + P7 + P8 + P9 = 1.
In this work, a brief analysis on how all the Pk relate to each other will be
done for each function that will be discussed. A particular method will be used
to derive a formula of the f dp of each function. As it was mentioned above, a

specific interval of the y-axis will be used as a range for each function, which is
[1, ∞). In addition, a certain partition of the interval into sections was proposed,
which breaks it into subintervals [10n−1 , 10n ) or [1 · 10n−1 , 10 · 10n−1 ), where n
is a natural number.
Each interval [10n−1 , 10 · 10n−1 ) will be partitioned into nine subintervals of
equal length, each of which can be expressed as [k · 10n−1 , (k + 1) · 10n−1 ), as
shown in Figure 5. Picking any number from the mentioned above subinterval
will automatically result on having a digit k as the leading digit of the number.
Fig 5. The specific measuring ruler for the range [10n−1 , 10n ).
Let us use a general real-valued continuous function y = f (x) on the range

[1, ∞) for the purpose of finding its f dp. We will derive a formula for Pk , which
will show the total probability of getting any y-value that starts with a digit k. In
other words, we need to find a formula for calculating the ratio of the probability
of a particular y-value being located in the interval [k·10n−1 , (k+1)·10n−1 ) to the
probability of the same value being located in the interval [1 · 10n−1 , 10 · 10n−1 )
or [10n−1 , 10n ) which is nine times longer.
Let us first find the formulas for Pk of each function on the above interval
and then we can analyze all the results.
The part of the continuous function y = f (x) that is shown in Figure 6
has the range [1 · 10n−1 , 10 · 10n−1 ). Therefore, the domain of this part is
[f −1 (1 · 10n−1 ), f −1 (10 · 10n−1 )). Moreover, all the x-values of the interval
[f −1 (1 · 10n−1 ), f −1 (10 · 10n−1 )) are mapped into the y-values of the interval
[k · 10n−1 , (k + 1) · 10n−1 ).
Also, all nine subintervals of the y-axis interval [1 · 10n−1 , 10 · 10n−1 ) have
equal length. However, nine corresponding subintervals of the x-axis interval
[f −1 (1 · 10n−1 ), f −1 (10 · 10n−1 )) are expected to have equal length only in a case
if y = f (x) is a linear function. In all other cases, unless a coincidence takes
place, all nine subintervals of this interval will have different length.
The percentage of the y-values mapped into [k · 10n−1 , (k + 1) · 10n−1 ) out
of those that mapped into [1 · 10n−1 , 10 · 10n−1 ) depends directly on the length
of the interval [f −1 (k · 10n−1 ), f −1 ((k + 1) · 10n−1 )). The longer this interval
is, the more numbers out of it will be mapped into the corresponding interval
[k · 10n−1 , (k + 1) · 10n−1 ).
The ratio of the length of the interval [f −1 (k · 10n−1 ), f −1 ((k + 1) · 10n−1 ))
to the interval [f −1 (1 · 10n−1 ), f −1 (10 · 10n−1 )) equals to the ratio of a number
of the y-values mapped into [k · 10n−1 , (k + 1) · 10n−1 ) to a number of those that
mapped into [1 · 10n−1 , 10 · 10n−1 ).
Fig 6. Continuous function y = f (x).
In other words,
f −1 ((k + 1) · 10n−1 ) − f −1 (k · 10n−1 )

Pk = (1)
f −1 (10 · 10n−1 ) − f −1 (1 · 10n−1 )
or
f −1 ((k + 1) · 10n−1 ) − f −1 (k · 10n−1 )
Pk = (2)
f −1 (10n ) − f −1 (10n−1 )
where k is a digit from 1 to 9 and n is a natural number. Since formula 2 is

shorter, it will be used in this work to calculate f dp of every continuous real-
valued function that will be discussed later.
2. Fdp of basic continuous functions
2.1. Pk of the exponential function y = ax
Let us first recall what the exponential function y = 2x looks like.
Fig 7. Exponential function y = 2x on the range [1, 10).
The first digit probabilities of the function, that is in the range [1, 10), are
explained in Figure 7. The ruler that was proposed earlier is used to measure
the probabilities.
As we can see clearly from the graph, all nine digits have different probabil-
ities in this range. After all the horizontal lines connecting the graph with its
corresponding y-values on the particular range were included to the picture, the
continuous exponential function might look like a discrete one with x-axis steps
of 0.2. Therefore, the conclusion that smaller digits have higher probabilities
can be proposed, but not finalized.
For instance, it might look like the digit 6 has a higher probability than the
digit 5. However, it is not true. The problem is that steps of 0.2 on that partic-
ular range are too big to show that in reality the smaller the digit is, the higher
its probability. That is why the formula for Pk is needed.

Before we derive the formula, let us look at other graphs of the exponential
function. Figure 8 shows the function y = 2x on the range [1, 1000). As we
can see, all the digits have different probabilities on that range as well as on
the previous one. However, the x-axis steps are not small enough to make any
specific conclusion about all Pk values on that range.
Let us look at the exponential function y = 3x now. Figure 9 shows the
function on the range [1, 10). Again, all the digits have different probabilities in
that interval.
The same function is shown on the range [1, 1000) in Figure 10. The new
range will bring us to the same conclusion: different digits have different prob-
abilities in this interval as well.
We had a chance to see the exponential functions with different bases on

different ranges. Let us derive the Pk formula now.
As it was mentioned above, the formula 2 will be used for each particular
function. We will derive the Pk formula for the exponential function y = ax
where 0 ≤ x < ∞ and a > 1. First, f (x) = ax . Then, f −1 (x) = loga x according
to the definition of the logarithmic function. Next,
n−1
loga ((k + 1) · 10n−1 ) − loga (k · 10n−1 ) loga (k+1)·10
k·10n−1
Pk = = n
loga 10n − loga 10n−1 loga 1010n−1
log10 k+1
loga k+1
k
k
log10 a log10 k+1
k log10 k+1
k k+1
= = log10 10
= = = log10 ;
loga 10 log10 10 1 k
log10 a
Thus,
k+1
Pk = log10 (3)
k
The formula satisfies Benford’s law [7]. It tells us two very important facts.
First, the formula does not contain the exponent n as well as any other
references to the range. It means that all nine Pk values of the exponential
function will be the same on any range.
Second, the formula does not have any reference to the initial general base a
of the function. It means that all Pk values do not depend on the base of the
exponential function. In other words, all exponential functions with the base
a > 1(y = 2x , y = ex , y = 3x , etc.) have the same f dp values.
Due to that fact that k is a whole number from 1 to 9, the newly derived
function should be considered as a discrete one on the domain [1, 9]. The graph
of it is shown in Figure 11.
Now, it is clear that the smaller first digit is, the higher its probability. More-
over, all the probabilities are positive numbers less than 1. This fact satisfies
the probability rules.
The only thing that we still need to check is whether the sum of all nine
probabilities will total to 1. A calculator could definitely be used for this purpose;
however, there is a couple of more neat solutions to the problem.
First, let us modify the formula.
k+1
Pk = log10 = log10 (k + 1) − log10 k (4)
k
Then, we will calculate the sum:
k+1
Fig 11. Function Pk = log10 k
.
Σ9k=1 Pk = Σ9k=1 (log10 (k + 1) − log10 k)
= (log10 (1 + 1) − log10 1) + (log10 (2 + 1) − log10 2) + (log10 (3 + 1) − log10 3)
+(log10 (4 + 1) − log10 4) + (log10 (5 + 1) − log10 5) + (log10 (6 + 1) − log10 6)
+(log10 (7 + 1) − log10 7) + (log10 (8 + 1) − log10 8) + (log10 (9 + 1) − log10 9)
= log10 2 − log10 1 + log10 3 − log10 2 + log10 4 − log10 3
+ log10 5 − log10 4 + log10 6 − log10 5 + log10 7 − log10 6
+ log10 8 − log10 7 + log10 9 − log10 8 + log10 10 − log10 9
= log10 10 − log10 1 = 1 − 0 = 1;
Thus, Σ9k=1 Pk = 1, so we got one of the proofs.

Then, let us graph the function y = log10 x on the domain from 1 to 10.
Figure 12 shows the graph.
Fig 12. Function y = log10 x.

The graph can be used to find the values of log10 k and log10 (k + 1) where
k is a whole number from 1 to 9. Both values can be found on the y-axis. After
picking a particular k, subtracting log10 k from log10 (k + 1) will give us the Pk
value according to the formula 4. The graph shows the differences between the
values log10 (k + 1) and log10 k for all nine values of k. All the differences are
equal to their corresponding Pk values. They are labeled with the corresponding
colors that are taken from the ruler described above.
The graph explains that the sum of all nine Pk values of the exponential
function total to 1. It is another proof of the fact that satisfies one more condition
of probabilities that we were looking for earlier.
Let us now calculate the Pk values of the exponential function and list them
in Table 1. As it was mentioned above, all the Pk values are positive numbers
less than 1 and they total to the number 1. In addition, smaller k values have
higher Pk values.
Table 1
Pk values of the exponential function
Digit Pk for y = ax
1 0.30103000
2 0.17609126
3 0.12493874
4 0.09691001
5 0.07918125
6 0.06694679
7 0.05799195
8 0.05115252
9 0.04575749
Sum 1.00000000
Let us think now about how the exponential function can model a real life
situation. We know that there is a group of problems like compound interest,
population growth, bacteria growth, and others that could be used to fulfill
our needs [5]. However, the bacteria growth problem will be the most suitable
for us due to that fact that bacteria grow much faster than money in a bank
or population in a particular area. Dealing with an exponential function, which
grows too slow, would prevent us from getting accurate results for the Pk values,
unless we want to analyze our function during an incredibly long time interval.
Thus, we are talking about bacterial growth. The formula that will be used
is:
N (t) = N0 ert (5)
N (t) - Final Amount of Bacteria

N0 - Initial Amount of Bacteria
e ≈ 2.71
r - Rate of Bacterial Growth
t - Time in Hours
We have discovered so far that the function y = ax where a > 1 has Benford’s
distribution. However, this function does not have any additional coefficient N0
and its exponent does not have any additional coefficient r as well. Therefore,
extra work needs to be done in order to make any conclusion about the f dp of
the bacterial growth function.
Let us figure out how the facts of multiplying the variable x by its coeffi-
cient and multiplying the entire function by another coefficient changes the f dp
function.
Let y = hamx where h, a, and m are constants and h > 0, a > 1, m > 0.
Next, we will switch x and y.
x = hamy ; x
h = amy ; my = loga ( hx ); y = 1
m loga ( hx ); f (x)−1 = 1
m loga ( hx )
We will use the same technique for finding the Pk values for the function
y = hamx that was used before to find the Pk values for the y = ax function.
Thus, we will start with Formula 2.
f −1 ((k + 1) · 10n−1 ) − f −1 (k · 10n−1 )

Pk =
f −1 (10n ) − f −1 (10n−1 )
1 (k+1)·10n−1 1 n−1
m loga h −m loga k·10h

= 1 10n 1 10n−1
m loga h − m loga h
1 (k+1)·10n−1 n−1
m (loga h − loga k·10h )

= 1 10n 10n−1
m (loga h − loga h )
(k+1)·10n−1 n−1
loga h − loga k·10h
= n n−1
loga 10h − loga 10 h
loga ((k + 1) · 10n−1 ) − loga h − loga (k · 10n−1 ) + loga h
=
loga (10n ) − loga h − loga (10n−1 ) + loga h
loga ((k + 1) · 10n−1 ) − loga (k · 10n−1 )
= ;
loga (10n ) − loga (10n−1 )
The last fraction shown above was used as the first step for getting the
Benford’s formula. It means that if we repeat the same steps as we already have
done, we will derive the exact same Formula 3 for our function y = hamx . Thus,
both constants h and m have no effects on f dp of the exponential function
y = hamx . The Pk values will stay the same which means they will satisfy
Benford’s law [7].
Fig 13. Function N (t) = 300e0.04t .
Let us come back to the formula N (t) = N0 ert . We should assign values to
N0 and r. For our bacterial growth problem, let the initial number of bacteria
be 300 and the rate of growth be 40%. Thus, N0 = 300 and r = 0.4. It means
that N (t) = 300e0.04t .
Let us graph the function and then we can do some calculations. The graph
describing bacterial growth during first twenty hours is shown in Figure 13.
If we want to calculate the number of bacteria regularly and to be able to
make a correct conclusion about the numbers, we need to do the calculations
with the same probability. We will choose the time interval of one hour and
calculate the number of bacteria for the first 200 hours. At least that many
calculations are needed because we need to analyze as many as nine different
probabilities. If we choose a relatively small number of time intervals, we will
prevent ourselves from getting accurate probabilities.
Let us use the formula N (t) = 300e0.4t and record all the calculations of the
value N (t) into Table 2. Even though the caluclated N (t) values are approxi-
mated, they all definitely have the correct first digits.
In order to make our results fair, we will not include the last two values to
our calculations because both of them contain numbers that start a new section
in the range and that section is not fully represented.
Table 2
Calculated N (t) values and their first digits
Number of Amount of First Number of Amount of First

Hours Bacteria Digit Hours Bacteria Digit
1 4.48E+02 4 51 2.17E+11 2
2 6.68E+02 6 52 3.24E+11 3
3 9.96E+02 9 53 4.83E+11 4
4 1.49E+03 1 54 7.21E+11 7
5 2.22E+03 2 55 1.08E+12 1
6 3.31E+03 3 56 1.60E+12 1
7 4.93E+03 4 57 2.39E+12 2
8 7.36E+03 7 58 3.57E+12 3
9 1.10E+04 1 59 5.33E+12 5
10 1.64E+04 1 60 7.95E+12 7
11 2.44E+04 2 61 1.19E+13 1
12 3.65E+04 3 62 1.77E+13 1
13 5.44E+04 5 63 2.64E+13 2
14 8.11E+04 8 64 3.94E+13 3
15 1.21E+05 1 65 5.87E+13 5
16 1.81E+05 1 66 8.76E+13 8
17 2.69E+05 2 67 1.31E+14 1
18 4.02E+05 4 68 1.95E+14 1
19 5.99E+05 5 69 2.91E+14 2
20 8.94E+05 8 70 4.34E+14 4
21 1.33E+06 1 71 6.47E+14 6
22 1.99E+06 1 72 9.66E+14 9
23 2.97E+06 2 73 1.44E+15 1
24 4.43E+06 4 74 2.15E+15 2
25 6.61E+06 6 75 3.21E+15 3
26 9.86E+06 9 76 4.78E+15 4
27 1.47E+07 1 77 7.13E+15 7
28 2.19E+07 2 78 1.06E+16 1
29 3.27E+07 3 79 1.59E+16 1
30 4.88E+07 4 80 2.37E+16 2
31 7.28E+07 7 81 3.53E+16 3
32 1.09E+08 1 82 5.27E+16 5
33 1.62E+08 1 83 7.86E+16 7
34 2.42E+08 2 84 1.17E+17 1
35 3.61E+08 3 85 1.75E+17 1
36 5.38E+08 5 86 2.61E+17 2
37 8.03E+08 8 87 3.90E+17 3
38 1.20E+09 1 88 5.81E+17 5
39 1.79E+09 1 89 8.67E+17 8
40 2.67E+09 2 90 1.29E+18 1
41 3.98E+09 3 91 1.93E+18 1
42 5.93E+09 5 92 2.88E+18 2
43 8.85E+09 8 93 4.29E+18 4
44 1.32E+10 1 94 6.41E+18 6
45 1.97E+10 1 95 9.56E+18 9
46 2.94E+10 2 96 1.43E+19 1
47 4.38E+10 4 97 2.13E+19 2
48 6.54E+10 6 98 3.17E+19 3
49 9.76E+10 9 99 4.73E+19 4
50 1.46E+11 1 100 7.06E+19 7
Number of Amount of First Number of Amount of First

Hours Bacteria Digit Hours Bacteria Digit
101 1.05E+20 1 151 5.11E+28 5
102 1.57E+20 1 152 7.62E+28 7
103 2.34E+20 2 153 1.14E+29 1
104 3.50E+20 3 154 1.70E+29 1
105 5.22E+20 5 155 2.53E+29 2
106 7.78E+20 7 156 3.78E+29 3
107 1.16E+21 1 157 5.63E+29 5
108 1.73E+21 1 158 8.40E+29 8
109 2.58E+21 2 159 1.25E+30 1
110 3.86E+21 3 160 1.87E+30 1
111 5.75E+21 5 161 2.79E+30 2
112 8.58E+21 8 162 4.16E+30 4
113 1.28E+22 1 163 6.21E+30 6
114 1.91E+22 1 164 9.26E+30 9
115 2.85E+22 2 165 1.38E+31 1
116 4.25E+22 4 166 2.06E+31 2
117 6.34E+22 6 167 3.08E+31 3
118 9.46E+22 9 168 4.59E+31 4
119 1.41E+23 1 169 6.85E+31 6
120 2.11E+23 2 170 1.02E+32 1
121 3.14E+23 3 171 1.52E+32 1
122 4.68E+23 4 172 2.27E+32 2
123 6.99E+23 6 173 3.39E+32 3
124 1.04E+24 1 174 5.06E+32 5
125 1.56E+24 1 175 7.55E+32 7
126 2.32E+24 2 176 1.13E+33 1
127 3.46E+24 3 177 1.68E+33 1
128 5.16E+24 5 178 2.51E+33 2
129 7.70E+24 7 179 3.74E+33 3
130 1.15E+25 1 180 5.58E+33 5
131 1.71E+25 1 181 8.32E+33 8
132 2.56E+25 2 182 1.24E+34 1
133 3.82E+25 3 183 1.85E+34 1
134 5.69E+25 5 184 2.76E+34 2
135 8.49E+25 8 185 4.12E+34 4
136 1.27E+26 1 186 6.15E+34 6
137 1.89E+26 1 187 9.17E+34 9
138 2.82E+26 2 188 1.37E+35 1
139 4.21E+26 4 189 2.04E+35 2
140 6.27E+26 6 190 3.04E+35 3
141 9.36E+26 9 191 4.54E+35 4
142 1.40E+27 1 192 6.78E+35 6
143 2.08E+27 2 193 1.01E+36 1
144 3.11E+27 3 194 1.51E+36 1
145 4.64E+27 4 195 2.25E+36 2
146 6.92E+27 6 196 3.36E+36 3
147 1.03E+28 1 197 5.01E+36 5
148 1.54E+28 1 198 7.47E+36 7
149 2.30E+28 2 199 1.11E+37
150 3.43E+28 3 200 1.66E+37
Let us calculate the first digit probabilities now. Table 3 has the analysis,
which was done using the table above and Benford’s probabilities for making
a comparison. As we can see, the probabilities from our table are very close to
those of Benford. We can also check that the sum of all totals is 198 and the
sum of each column with probabilities is equal to 1.
Table 3
Analysis of N (t) values’ first digits
Digit Count Frequency Benford’s

(Count / Sum) Pk values
1 59 0.29797980 0.30103000
2 34 0.17171717 0.17609126
3 25 0.12626263 0.12493874
4 19 0.09595960 0.09691001
5 17 0.08585859 0.07918125
6 13 0.06565657 0.06694679
7 12 0.06060606 0.05799195
8 10 0.05050505 0.05115252
9 9 0.04545455 0.04575749
Sum 198 1.00000000 1.00000000
2.2. Pk of the power function y = xa
Let us talk about the function y = xa where a is a natural number greater than 1.
Fig 14. Quadratic function y = x2 on the range [1, 100).

Again, we will look at the function on the range [1, ∞) for the reasons ex-
plained in the introduction. In order to choose the correct domain, we will pick
the interval [1, ∞), so it will match the range mentioned above for this func-
tion. As we know, the functions of that kind with even exponents have shapes
different from shapes of those with odd exponents. However, they all look more
similar to each other on the domain that we will use for all of the functions
y = xa .
First, let a = 2. The graph of the function y = x2 on the limited range [1, 100)
is shown in Figure 14. The ruler that was used earlier is used again to measure
the probabilities.
According to the graph, all nine digits have different probabilities in this
range, which is correct. However, the continuous quadratic function was ap-
proximated to the discrete one with x-axis steps of one unit. That is why more
work needs to be done before we can prove that the smaller digits have higher
probabilities than the bigger ones. For instance, it might look like the digit 8
has a higher probability than the digit 7 and so on. However, this is not correct.
The reason is that the x-axis steps of one unit for this particular domain and
range are too big to show that the smaller the digit is, the higher its probability.
Fig 15. Quadratic function y = x2 on the range [1, 1000).

Let us look at another graph of the function y = x2 . Figure 15 shows this

function on the range [1, 1000). As we can see, all the digits have different
probabilities on that range as well as on the previous one. Unfortunately, the x-
axis steps are still not small enough to allow us to make any specific conclusion
about all Pk values on that range as well.
Fig 16. Cubic function y = x3 on the range [1, 10).
Fig 17. Cubic function y = x3 on the range [1, 1000).

Let us look now at the function y = x3 now. Figure 16 shows the function
on the range [1, 10). Again, all the digits have different probabilities in that
interval.
The same function is shown on the range [1, 1000) in Figure 17. The new
range will provide us with the same conclusion: different digits have different
probabilities in this interval as well.
After we saw the functions y = xa with different exponents on different
ranges, let us derive the Pk formula for the function.
The Formula 2 should be used to accomplish our goal. Let us also remind
ourselves the restrictions: 1 ≤ x < ∞ and a is a natural number greater than 1.
Let f (x) = xa . This function is a one-to-one function on the domain [0, ∞)
regardless of whether the exponent a is even or odd. Thus, the function has an
1
inverse, which is f −1 (x) = x a . Next,
1 1
((k + 1) · 10n−1 ) a − (k · 10n−1 ) a
Pk = 1 1
(10n ) a − (10n−1 ) a
1 1 1 1 1
(10n−1 ) a ((k + 1) a − k a ) (k + 1) a − k a
= 1 1 = 1 ;
(10n−1 ) a (10 a − 1) 10 a − 1
Thus,
1 1
(k + 1) a − k a
Pk = 1 (6)
10 a − 1
Let us analyze the result. As we can see, the final formula does not contain
the exponent n or any other references to the functions range. It means that the
function will have the same f dp on any range. However, the formula contains the
exponent a. Therefore, we can conclude that the f dp of the quadratic function
y = x2 is different from the f dp of the cubic function y = x3 or any other
function of that nature that has a different exponent.
Let us prove that the sum of all the Pk of the function equals to 1.
1 1
−k a
Σ9k=1 Pk = Σ9k=1 (k+1)1
a
10 a −1
1 1 1 1 1 1 1 1 1 1 1 1
(1+1) a −1 a −2 a −3 a −4 a −5 a −6 a
+ (2+1)1 + (3+1)1 + (4+1)1 + (5+1)1 + (6+1)1
a a a a a
= 1
10 a −1 10 a −1 10 a −1 10 a −1 10 a −1 10 a −1
1 1 1 1 1 1
(7+1) a −7 a (8+1) a −8 a (9+1) a −9 a
+ 1 + 1 + 1
10 a −1 10 a −1 10 a −1
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
2 a −1 a +3 a −2 a +4 a −3 a +5 a −4 a +6 a −5 a +7 a −6 a +8 a −7 a +9 a −8 a +10 a −9 a
= 1
10 a −1
1 1 1
10 a −1 a 10 a −1
= 1 = 1 = 1;
10 a −1 10 a −1
Thus, Σ9k=1 Pk = 1, which satisfies one of the probabilitys law.

Since the probabilities of both functions y = x2 and y = x3 are analyzed in
this work, let us now derive the Pk formulas for each of them.
First, let y = x2 . Then a = 2 and

1 1 √ √ √ √
(k + 1) 2 − k 2 k+1− k k+1− k
Pk = 1 = √ ≈ (7)
10 2 − 1 10 − 1 2.16
Next, let y = x3 . Then a = 3 and
1 1 √ √ √ √
(k + 1) 3 − k 3 3
k+1− 3k 3
k+1− 3k
Pk = 1 = √ ≈ (8)
10 3 − 1 10 − 1 1.15
Let us recall that k is a whole number from 1 to 9. As it was done in the previous
section, we should graph both functions derived above as discrete ones on the
domain [1, 9]. Both graphs are shown in Figure 18.
The graphs show that the smaller the first digit is, the higher its probability
and that all the probabilities are positive numbers less than 1. The last fact
satisfies the probability rules.
Fig 18. Pk values of the functions y = x2 and y = x3 .
We already had a chance to check whether the sum of all nine probabilities
totals to 1. Let us find one more way to prove the same fact; however, we will
use graphs this time.
Since the Formula 6 tells us that the value of Pk depends on the exponent a,
we should sketch one graph proving that Σ9k=1 Pk = 1 for the function y = x2
and another graph proving that Σ9k=1 Pk = 1 for the function y = x3 . √
x
First, let a = 2. We will use the Formula 7. The function f (x) = √10−1 on
the domain [1, 10] is shown in Figure 19.
√
x
Fig 19. The function f (x) = √ .
10−1
√ √
x x+1
The graph can be used to find the values of √10−1 and √10−1 where k is a
whole number from 1 to 9. Both of those values √
can be√ found on the y-axis.
x x+1
After picking a particular k, subtracting √10−1 from √10−1 will give us the
corresponding
√
P k value.√
The graph in Figure 19 shows the differences between
x+1 x
the values 10−1 and 10−1 for all nine values of k. All the differences are equal
√ √
to the corresponding Pk values. They are labeled with the corresponding colors
that are taken from the ruler described above.
In addition, the graph clearly shows that the sum √
of all nine
√
Pk values
√
of the
10 1
function equals to the difference f (10) − f (1) = √10−1 − √10−1 = √10−1
10−1
= 1.
Thus, we have proved once again that Σk=1 Pk = 1 for the function y = x2 .
9
Let us now do the same for the function y = x3 . √

3
x
Let a = 3. the Formula 8 will be used. The function f (x) = √ 3
10−1
on the
domain [1, 10] is shown in Figure 20. Using the same technique that was used for
the square function, we can see that the sum of all√nine Pk √ values of√ the cubic
3 3 3
10 1 10−1
function equals to the difference f (10) − f (1) = √3
10−1
− √
3
10−1
= √ 3
10−1
= 1.
9 3
It proves once again that Σk=1 Pk = 1 for the function y = x .
Let us now calculate the Pk values of the quadratic and cubic functions.
They are listed in Table 4. As we can see, all the Pk values for both functions
are positive numbers less than 1 and they total to the number 1. Moreover,
smaller k values have higher Pk values.
√
3x
Fig 20. The function f (x) = √
3 .
10−1
Table 4
Pk values for the quadratic and cubic functions
Digit Pk for y = x2 Pk for y = x3
1 0.19156354 0.22515007
2 0.14699187 0.15793749
3 0.12391988 0.12573382
4 0.10917561 0.10617742
5 0.09870229 0.09281135
6 0.09076613 0.08299351
7 0.08448305 0.07542117
8 0.07934822 0.0693706
9 0.07504941 0.06440457
Sum 1.00000000 1.00000000
Before we talk about how the quadratic, cubic or other functions of this kind
can model a real life situation, we should check how a coefficient in front of the
variable x and another one in front of the whole function would affect the Pk
values of the function. We will do them one by one by considering two cases:
y = (mx)a and y = hxa where h and m are positive constants.
First, let y = (mx)a . Then, y = ma xa . As we can see, ma is a coefficient in
front of the whole function and it falls into the y = hxa category. Therefore, we
do not need to analyze them separately. Let us just talk about y = hxa because
it looks easier.
Thus, y = hxa . Next, we will switch x and y:
1 1
x = hy a ; hx = y a ; y = ( hx ) a ; Let f (x)−1 = ( hx ) a . Then,
n−1 1 n−1 1
f −1 ((k + 1) · 10n−1 ) − f −1 (k · 10n−1 ) ( (k+1)·10
h ) a − ( k·10h ) a
Pk = = n−1 1 n−1 1
f −1 (10 · 10n−1 ) − f −1 (1 · 10n−1 ) ( 10·10h ) a − ( 1·10h ) a
n−1 1 1 1 1 1
(( 10 h ) a )((k + 1) a − k a ) (k + 1) a − k a
= n−1 1 1 = 1 ;
(( 10 h ) a )(10 a − 1) 10 a − 1
In conclusion, we came back to the Formula 6. It means that the Pk values

of the function y = hxa are the same as the Pk values of the function y = xa .
We are ready to talk about an actual real-life problem [5]. Let us look at the
motion of an object, released from a height of 1000 feet. If we disregard the air
resistance, the distance of the object from the initial point can be calculated
by a formula D(t) = 21 gt2 , where g = 32.2f t/s2 and t is time in seconds. Thus,
D(t) = 16.1t2 . It means that we will be dealing with the function y = hx2 ,
where y = D, h = 16.1, and x = t. The graph in Figure 21 shows how the
distance is changing through the time.
Fig 21. The function D(t) = 16.1t2 .

Table 5
Calculated D(t) values and their first digits
Time in Distance First Time in Distance First Time in Distance First

Seconds in Feet Digit Seconds in Feet Digit Seconds in Feet Digit
0.05 0.04025 2.7 117.369 1 5.35 460.8222 4
0.1 0.161 2.75 121.7563 1 5.4 469.476 4
0.15 0.36225 2.8 126.224 1 5.45 478.2102 4
0.2 0.644 2.85 130.7723 1 5.5 487.025 4
0.25 1.00625 1 2.9 135.401 1 5.55 495.9202 4
0.3 1.449 1 2.95 140.1103 1 5.6 504.896 5
0.35 1.97225 1 3 144.9 1 5.65 513.9522 5
0.4 2.576 2 3.05 149.7703 1 5.7 523.089 5
0.45 3.26025 3 3.1 154.721 1 5.75 532.3062 5
0.5 4.025 4 3.15 159.7523 1 5.8 541.604 5
0.55 4.87025 4 3.2 164.864 1 5.85 550.9822 5
0.6 5.796 5 3.25 170.0563 1 5.9 560.441 5
0.65 6.80225 6 3.3 175.329 1 5.95 569.9802 5
0.7 7.889 7 3.35 180.6823 1 6 579.6 5
0.75 9.05625 9 3.4 186.116 1 6.05 589.3002 5
0.8 10.304 1 3.45 191.6303 1 6.1 599.081 5
0.85 11.63225 1 3.5 197.225 1 6.15 608.9422 6
0.9 13.041 1 3.55 202.9002 2 6.2 618.884 6
0.95 14.53025 1 3.6 208.656 2 6.25 628.9062 6
1 16.1 1 3.65 214.4922 2 6.3 639.009 6
1.05 17.75025 1 3.7 220.409 2 6.35 649.1922 6
1.1 19.481 1 3.75 226.4062 2 6.4 659.456 6
1.15 21.29225 2 3.8 232.484 2 6.45 669.8002 6
1.2 23.184 2 3.85 238.6422 2 6.5 680.225 6
1.25 25.15625 2 3.9 244.881 2 6.55 690.7302 6
1.3 27.209 2 3.95 251.2002 2 6.6 701.316 7
1.35 29.34225 2 4 257.6 2 6.65 711.9822 7
1.4 31.556 3 4.05 264.0802 2 6.7 722.729 7
1.45 33.85025 3 4.1 270.641 2 6.75 733.5562 7
1.5 36.225 3 4.15 277.2822 2 6.8 744.464 7
1.55 38.68025 3 4.2 284.004 2 6.85 755.4522 7
1.6 41.216 4 4.25 290.8062 2 6.9 766.521 7
1.65 43.83225 4 4.3 297.689 2 6.95 777.6702 7
1.7 46.529 4 4.35 304.6522 3 7 788.9 7
1.75 49.30625 4 4.4 311.696 3 7.05 800.2102 8
1.8 52.164 5 4.45 318.8202 3 7.1 811.601 8
1.85 55.10225 5 4.5 326.025 3 7.15 823.0722 8
1.9 58.121 5 4.55 333.3102 3 7.2 834.624 8
1.95 61.22025 6 4.6 340.676 3 7.25 846.2562 8
2 64.4 6 4.65 348.1222 3 7.3 857.969 8
2.05 67.66025 6 4.7 355.649 3 7.35 869.7622 8
2.1 71.001 7 4.75 363.2562 3 7.4 881.636 8
2.15 74.42225 7 4.8 370.944 3 7.45 893.5902 8
2.2 77.924 7 4.85 378.7122 3 7.5 905.625 9
2.25 81.50625 8 4.9 386.561 3 7.55 917.7402 9
2.3 85.169 8 4.95 394.4902 3 7.6 929.936 9
2.35 88.91225 8 5 402.5 4 7.65 942.2122 9
2.4 92.736 9 5.05 410.5902 4 7.7 954.569 9
2.45 96.64025 9 5.1 418.761 4 7.75 967.0062 9
2.5 100.625 1 5.15 427.0122 4 7.8 979.524 9
2.55 104.6903 1 5.2 435.344 4 7.85 992.1222 9
2.6 108.836 1 5.25 443.7562 4 7.9 1004.801
2.65 113.0623 1 5.3 452.249 4
Let us calculate how many seconds it will take the object to reach the ground
after falling from 1000 feet height. First, 1000 = 16.1t2 . Then, t2 ≈ 62.1. Thus,
t ≈ 7.9. Since the object will touch the ground in almost 8 seconds, we will
calculate the distance of an object from the initial point every twentieth portion
of a second. It will give us enough numbers for our table.
Let us look at Table 5. The first column shows how many seconds have passed
after the object was released. The second column shows the distance that the
object have gone through after the corresponding time period have passed. The
third column shows the first digit of the number representing the distance.
Since four first values in the distance column do not belong to the range of
[1, ∞) that we normally use, we will not do anything about the first digits of
these values. In addition, the last calculated distance of 1004.801 exceeds the
limit of 1000 feet that is why we will not do anything about the first digit of
this number as well.
Now, let us analyze the first digit probabilities. Table 5 calculations and
the quadratic function probabilities for a comparison are shown in Table 6. The
calculated free fall probabilities are very close to those of the quadratic function.
If we had calculated the distance of the object more frequently than we did, we
would get probabilities that are more accurate.
Table 6
Analysis of D(t) values’ first digits
Digit Count Frequency Pk values

(Count / Sum) for y = x2
1 31 0.20261438 0.19156354
2 22 0.14379085 0.14699187
3 18 0.11764706 0.12391988
4 18 0.11764706 0.10917561
5 15 0.09803922 0.09870229
6 13 0.08496732 0.09076613
7 13 0.08496732 0.08448305
8 12 0.07843137 0.07934822
9 11 0.07189542 0.07504941
Sum 153 1.00000000 1.00000000
2.3. Pk of the linear function y = mx
Let us look at the linear function y = mx where m > 0. The reason why we are
choosing a positive slope is that fact that our initial goal is to apply each function
in this work to a real life situation. We should admit that a graph reflecting any
such situation would take the first quadrant only. However, choosing a negative
slope will prevent the graph from passing through the first quadrant. Thus, we
are choosing m > 0.
In addition, the slope cannot be equal to zero because otherwise we would
get a horizontal line y = 0, which would prevent us from talking about any
probability at all. All the first digits in this case would just be zeroes.
We will use that portion of the function, which does not go beyond the
range [1, ∞) since we agreed upon using this range for all our functions in the
introduction to the work. A particular value of m and the previously mentioned
1
range will lead us to have the domain [ m , ∞).
Let us look at Figure 22, which shows a graph of a linear function y = 5x
in the limited range from 1 to 100. As we can see, the graphs direction will not
be changed if we make the range smaller or bigger. Moreover, according to the
colors of all the horizontal lines, an assumption that all the digits have equal
probabilities can be made. We just have to prove it.
Fig 22. The function y = 5x.
The formula 2 will be used. Let f (x) = mx. Then, f −1 (x) = 1

m x. Next,
1 1
m · (k + 1) · 10n−1 − m · k · 10n−1
Pk = 1 n 1 n−1
m · 10 − m · 10
1
m · 10n−1 ((k + 1) − k) k+1−k 1
= 1 = = ;
m · 10
n−1 (10 − 1) 10 − 1 9
Thus,
1
Pk = (9)
9
The formula looks a bit unusual; however, it proves the assumption made by us
before: all nine digits have the same probabilities regardless of the slope of the
linear function or chosen range. In addition, it is very easy to check that the
total of all nine Pk values equals to 1. The formula derived above will lead us to
a horizontal line if we decide to graph it. Figure 23 shows the probability line
of a linear function.
Fig 23. Pk values of the function y = mx.
Let us think about how a linear function can be applied to a real life situation
[5]. Consider a swimming pool, which holds 1000f t3 of water. We will fill it with
water using a garden hose, which has a flow rate of 5f t3 /min. After the hose is
turned on, the volume of water in the pool before it is filled can be calculated
with the formula V (t) = 5t, where V (t) is the volume measured in f t3 and t is
time in minutes. It means that we will be dealing with the function y = mx,
where y = V (t), m = 5, and x = t. The graph in Figure 22 can be used to see
how the amount of water is increasing during the first 20 minutes.
Let us calculate how many minutes it will take to fill the entire pool with
water. Let 1000 = 5t. Then t = 1000/5 = 200. Since the pool will be filled in
200 minutes, we will calculate the amount of water in it every minute. Doing
this should give us enough numbers for our table.
Let us look at Table 7. The first column shows how many minutes have passed
after the hose was turned on. The second column shows the volume of water
in the pool in f t3 after the corresponding number of minutes have passed. The
third column shows the first digit of the number representing the volume.
Since the table has only one y-value in the range [1, 10), that value should
be excluded from our calculations to keep them fair. In addition, the last y-
value of our table exceeds the range of [1, 1000). Therefore, that value should
be excluded from our calculations as well.
Table 7
Calculated V (t) values and their first digits
Time in Volume First Time in Volume First Time in Volume First

Seconds in f t3 Digit Seconds in f t3 Digit Seconds in f t3 Digit
1 5 51 255 2 101 505 5
2 10 1 52 260 2 102 510 5
3 15 1 53 265 2 103 515 5
4 20 2 54 270 2 104 520 5
5 25 2 55 275 2 105 525 5
6 30 3 56 280 2 106 530 5
7 35 3 57 285 2 107 535 5
8 40 4 58 290 2 108 540 5
9 45 4 59 295 2 109 545 5
10 50 5 60 300 3 110 550 5
11 55 5 61 305 3 111 555 5
12 60 6 62 310 3 112 560 5
13 65 6 63 315 3 113 565 5
14 70 7 64 320 3 114 570 5
15 75 7 65 325 3 115 575 5
16 80 8 66 330 3 116 580 5
17 85 8 67 335 3 117 585 5
18 90 9 68 340 3 118 590 5
19 95 9 69 345 3 119 595 5
20 100 1 70 350 3 120 600 6
21 105 1 71 355 3 121 605 6
22 110 1 72 360 3 122 610 6
23 115 1 73 365 3 123 615 6
24 120 1 74 370 3 124 620 6
25 125 1 75 375 3 125 625 6
26 130 1 76 380 3 126 630 6
27 135 1 77 385 3 127 635 6
28 140 1 78 390 3 128 640 6
29 145 1 79 395 3 129 645 6
30 150 1 80 400 4 130 650 6
31 155 1 81 405 4 131 655 6
32 160 1 82 410 4 132 660 6
33 165 1 83 415 4 133 665 6
34 170 1 84 420 4 134 670 6
35 175 1 85 425 4 135 675 6
36 180 1 86 430 4 136 680 6
37 185 1 87 435 4 137 685 6
38 190 1 88 440 4 138 690 6
39 195 1 89 445 4 139 695 6
40 200 2 90 450 4 140 700 7
41 205 2 91 455 4 141 705 7
42 210 2 92 460 4 142 710 7
43 215 2 93 465 4 143 715 7
44 220 2 94 470 4 144 720 7
45 225 2 95 475 4 145 725 7
46 230 2 96 480 4 146 730 7
47 235 2 97 485 4 147 735 7
48 240 2 98 490 4 148 740 7
49 245 2 99 495 4 149 745 7
50 250 2 100 500 5 150 750 7
Time in Volume First Time in Volume First Time in Volume First

Seconds in f t3 Digit Seconds in f t3 Digit Seconds in f t3 Digit
151 755 7 168 840 8 185 925 9
152 760 7 169 845 8 186 930 9
153 765 7 170 850 8 187 935 9
154 770 7 171 855 8 188 940 9
155 775 7 172 860 8 189 945 9
156 780 7 173 865 8 190 950 9
157 785 7 174 870 8 191 955 9
158 790 7 175 875 8 192 960 9
159 795 7 176 880 8 193 965 9
160 800 8 177 885 8 194 970 9
161 805 8 178 890 8 195 975 9
162 810 8 179 895 8 196 980 9
163 815 8 180 900 9 197 985 9
164 820 8 181 905 9 198 990 9
165 825 8 182 910 9 199 995 9
166 830 8 183 915 9 200 1000
167 835 8 184 920 9
We should use the table above to analyze the first digit probabilities of the
numbers representing the volume of water. Table 8 shows the calculations from
the table above and the linear function probabilities for making a comparison.
As we can see, the calculated water volume f dp values are the same as the
corresponding f dp values of the liner function. Moreover, all the water volume
f dp values are equal to each other, as it should be.
Table 8
Analysis of V (t) values’ first digits
Digit Count Frequency Pk values

(Count / Sum) for y = mx
1 22 1/9 1/9
2 22 1/9 1/9
3 22 1/9 1/9
4 22 1/9 1/9
5 22 1/9 1/9
6 22 1/9 1/9
7 22 1/9 1/9
8 22 1/9 1/9
9 22 1/9 1/9
Sum 198 1.00000000 1.00000000
√
a
2.4. Pk of the root function y = x
√
Let us talk about the function y = x where a is a natural number greater
a
than 1. Again, we will look at the function on the range [1, ∞). Moreover, we
will pick the same interval [1, ∞) for the domain, so that they will both match
each other. √
First, let a = 2. The graph of the function y = x on the limited range
[1, 100) is shown in Figure 24. The same ruler is used to measure the probabili-
ties. Looking at the graph, we can see that bigger digits have higher probabilities.
However, more work needs to be done to prove it.
√
Fig 24. The function y = x.
√
3
Fig 25. The function y = x.
√
Let us look now at the function y = 3 x. Figure 25 shows the function on the
range [1, 100). Again, bigger digits have higher probabilities and we will prove
it. √
We had a chance to see the functions y = a x with different indexes a. Let
us derive the formula for Pk of the function.

The formula 2 should be used. In addition, we know that 1 ≤ x < ∞ and a
is a natural√number greater than 1.
Let y = a x. This function is a one-to-one function on any part of its domain
regardless of the index a. That is why the function has an inverse, which is
f −1 (x) = xa . Next,
((k + 1) · 10n−1 )a − (k · 10n−1 )a

Pk =
(10n )a − (10n−1 )a
(10n−1 )a ((k + 1)a − k a ) (k + 1)a − k a

= n−1 a a
= ;
(10 ) (10 − 1) 10a − 1
Consequently,
(k + 1)a − k a
Pk = (10)
10a − 1
The final formula does not contain the exponent n as well as any other references
to the functions range. It means that the function will have the same f dp on
any of its range. However, the formula contains the index a, which looks like an
exponent
√ now. Thus, we can conclude that the f dp of the square
√ root function
y = x is different from the f dp of the cube root function y = 3 x or any other
function of that nature that has a different index.
Let us prove that the sum of all the Pk of the function equals to 1.
a a
−k
Σ9k=1 Pk = Σ9k=1 (k+1)
10a −1
(1+1)a −1a (2+1)a −2a (3+1)a −3a (4+1)a −4a (5+1)a −5a (6+1)a −6a
= 10a −1 + 10a −1 + 10a −1 + 10a −1 + 10a −1 + 10a −1
a a
−7 (8+1)a −8a (9+1)a −9a
+ (7+1)
10a −1 + 10a −1 + 10a −1
2a −1a +3a −2a +4a −3a +5a −4a +6a −5a +7a −6a +8a −7a +9a −8a +10a −9a
= 10a −1
10a −1a 10a −1

= 10a −1 = 10a −1 = 1;
Therefore, Σ9k=1 Pk = 1, which satisfies one of the rules of probability.

Let us now derive the Pk formulas for the square root and the cube root
functions separately.
√
First, let y = x. Then a = 2 and
(k+1)2 −k2 k2 +2k+1−k2 2k+1
Pk = 102 −1 = 100−1 = 99 ;
√
Next, let y = 3
x. Then a = 3 and
(k+1)3 −k3 k3 +3k2 +3k+1−k3 3k2 +3k+1

Pk = 103 −1 = 1000−1 = 999 ;
Let us recall that k is a whole number from 1 to 9. We should graph both

functions derived above as discrete ones on the domain [1, 9]. The graphs of
both of them are shown in Figure 26.
2k+1 3k2 +3k+1

Fig 26. Functions Pk = 99
and Pk = 999
Both graphs show that the bigger the first digit is, the higher its probability
and that all the probabilities are positive numbers less than 1. The last fact
satisfies the probability rules.
Since we already checked that the sum of all nine probabilities totals to 1
using the formulas, let us prove the same fact using graphs this time.
Pk formula depends on the value of a, √ so we should sketch one graph proving
that Σ9k=1 Pk = 1 for the function
√ y = x and another graph proving the same
thing for the function y = 3 x.
2
−k2
First, let a = 2. The original formula Pk = (k+1) 99 should be used. Let
x2
us graph the function f (x) = 99 on the domain [1, 10]. The graph is shown in
Figure 27.
2 2
This graph can be used to find the values of (k+1) 99 and k99 where k is a
whole number from 1 to 9. Both of the values can be found on the y-axis. After
2 2
picking a particular k, subtracting k99 from (k+1)
99 will give us the corresponding
Pk value.
2
The graph in Figure 27 shows the differences between the values (k+1) 99 and
k2
99 for all nine values of k. All the differences are equal to the corresponding Pk
values. They are labeled with the corresponding colors that are taken from the
ruler described above.
x2
Fig 27. Functions f (x) = 99
on the domain [1, 10]
The graph clearly shows that the sum of all nine Pk values of the function
equals to the difference f (10) − f (1) = 100 1 99
99 − 99 = 99 = 1. √
9
We have proved once again that Σk=1 Pk = 1 for the function y = x.
x3
Fig 28. Functions f (x) = 999
√
Let us now do the same for the function y = 3 x. Let a = 3. The initial
3 3
−k x3
formula is Pk = (k+1) 999 . Let us graph the function f (x) = 999 on the domain
[1, 10]. The graph is shown in Figure 28.
Using the same technique that was used for the square root function, we
can see that the sum of all nine Pk values of the cubic function equals to the
difference f (10) − f (1) = 1000 1 999
999 − 999 = 999 = 1. √
9
It proves once again that Σk=1 Pk = 1 for the function y = 3 x.
Let us now look at the Pk values of the square root and the cube root func-
tions. They are listed in Table 9. As we can see, bigger k values have higher
Pk values in both cases. In addition, all the Pk values for both functions are
positive numbers less than 1. If we total the exact Pk values for each function
(not the approximated ones with three decimal places), the totals will be equal
to the number 1.
We also can see that for the digits from one to six, the fdp values of the square
root function are higher than those of the cube root function. However, the fdp
values of the digits from seven to nine are higher for the cube root function than
those of the square root function.
Table 9
Pk values for the square root and cube root functions
√ √
Digit Pk for y = x Pk for y = 3 x
1 0.03030303 0.00700701
2 0.05050505 0.01901902
3 0.07070707 0.03703704
4 0.09090909 0.06106106
5 0.11111111 0.09109109
6 0.13131313 0.12712713
7 0.15151515 0.16916917
8 0.17171717 0.21721722
9 0.19191919 0.27127127
Sum 1.00000000 1.00000000
√
Prior to talking about how the function y = a x can be used to model a real
life situation, let us check how a coefficient in front of the variable x and another
one in front of the whole function will affect the Pk values of the function.
√ Let
us consider
√ two cases similar to those that were done before: y = a
mx and
y = h a x where h√and m are positive √ constants.
√ √
First, let y = a mx. Then, y = a m · a x. Obviously, √
a
m is a coefficient in
front of the whole function and it falls into the y = h x category. Therefore,
a
we do not need√to analyze both functions separately.√Let us just talk about the
function y = h √ a
x because it looks easier than y = a mx.
Thus, y = h x. Next, we will switch x and y:
a
√ √
x = h a y; hx = a y; y = ( hx )a ; Let f (x)−1 = ( hx )a Then,
f −1 ((k + 1) · 10n−1 ) − f −1 (k · 10n−1 )

Pk =
f −1 (10 · 10n−1 ) − f −1 (1 · 10n−1 )
n−1 n−1
( (k+1)·10
h )a − ( k·10h )a
= 10·10n−1 10n−1
( h )a − ( h )a
n−1
( 10 h )a [(k + 1)a − k a ] (k + 1)a − k a
= 10n−1
= ;
( h )a (10a − 1) 10a − 1
a a
−k
Thus, we came back to the initial formula Pk = (k+1) 10a −1 . It means that the
√
Pk values
√ of the function y = h a x are the same as the Pk values of the function
y = a x.
Now, we are ready to talk about the actual task. Let us analyze the increase in
human height of young males. In particular, we will look at how a boys height
changes from birth to 18.5 years of age [6].
Let us look at the following table, which is a summary of a couple of original
tables published for public by the department of health and human services.
Table 10 shows the average height of boys from birth to 18.5 years of age.
Table 10
Average height of boys from birth to 18.5 years of age
Months Hight Months Hight Months Hight Months Hight
cm cm cm cm
0 51.40 15 77.81 84 120.19 162 158.49
1 52.04 18 80.64 90 123.47 168 160.98
2 55.54 21 83.75 96 126.61 174 166.13
3 58.73 24 86.33 102 128.62 180 168.42
4 62.09 30 89.18 108 131.58 186 170.61
5 64.15 36 93.47 114 134.71 192 172.39
6 66.68 42 96.82 120 136.91 198 173.31
7 68.49 48 100.45 126 139.59 204 175.63
8 70.17 54 104 132 142.32 210 175.78
9 71.69 60 107.14 138 144.65 216 176.1
10 72.52 66 110.94 144 147.9 222 177.53
11 73.18 72 113.89 150 151.43
12 74.95 78 117.21 156 154.79
Now, let us modify the table a little bit. First, as it was mentioned above,
we will not analyze the actual height, but the increase in height. So, the initial
height of 51.4 cm will be subtracted from each number.
Next, to make the calculation process of the f dp fair, the last y-value that
we are using should be close to a power of 10. Thus, 177.53 − 51.4 = 126.13.
1000
Then 126.13 ≈ 7.93 ≈ 8. That is why we should multiply each value in the height
increase column by 8. In other words, we will use a new unit of measure, which
equals to 1/8 cm. In this case, the last y-value will be approximately equal to
1000.
Let us look at Table 11. The first column is not changed and it provides us
with the x-values for our graph. The second column is not changed either. The
third column shows the difference in height by subtracting 51.4 from each value
in the second column. The last column contains the products of the values from
the previous column and number 8. This column provides us with the y-values
for our graph.
Table 11
Height growth of boys from birth to 18.5 years of age
Months Hight Difference y-values Months Hight Difference y-values
x-values cm cm x-values cm cm
0 51.40 0 0 78 117.21 65.81 526.48
1 52.04 0.64 5.12 84 120.19 68.79 550.32
2 55.54 4.14 33.12 90 123.47 72.07 576.56
3 58.73 7.33 58.64 96 126.61 75.21 601.68
4 62.09 10.69 85.52 102 128.62 77.22 617.76
5 64.15 12.75 102 108 131.58 80.18 641.44
6 66.68 15.28 122.24 114 134.71 83.31 666.48
7 68.49 17.09 136.72 120 136.91 85.51 684.08
8 70.17 18.77 150.16 126 139.59 88.19 705.52
9 71.69 20.29 162.32 132 142.32 90.92 727.36
10 72.52 21.12 168.96 138 144.65 93.25 746
11 73.18 21.78 174.24 144 147.9 96.5 772
12 74.95 23.55 188.4 150 151.43 100.03 800.24
15 77.81 26.41 211.28 156 154.79 103.39 827.12
18 80.64 29.24 233.92 162 158.49 107.09 856.72
21 83.75 32.35 258.8 168 160.98 109.58 876.64
24 86.33 34.93 279.44 174 166.13 114.73 917.84
30 89.18 37.78 302.24 180 168.42 117.02 936.16
36 93.47 42.07 336.56 186 170.61 119.21 953.68
42 96.82 45.42 363.36 192 172.39 120.99 967.92
48 100.45 49.05 392.4 198 173.31 121.91 975.28
54 104 52.6 420.8 204 175.63 124.23 993.84
60 107.14 55.74 445.92 210 175.78 124.38 995.04
66 110.94 59.54 476.32 216 176.1 124.7 997.6
72 113.89 62.49 499.92 222 177.53 126.13 1009.04
Now, we are ready to sketch a graph using the x and y-values from the above
table. Figure 29 shows the graph. The table provides us only with discrete values,
so to make the graph more smooth-looking we should draw a curve connecting
all the points.
As we
√ can see, the shape of the graph reminds us of the graph of the function
y = h a x. Moreover, if we do further analysis of the numbers in the table, √ we
will see that the closest function to the one graphed is the function y = h x.
The graph shows that the younger a boy is, the faster he grows. It also shows
that particular first digits have higher probabilities; however, we are not yet
ready to state that the bigger a digit is, the higher its probability. First, the
intervals between the x-values in the table are not equal to each other. That
problem will prevent us from getting the probabilities correctly, so it needs to
be fixed. In addition, the top portion of the graph has a couple of places where
the slope changes more rapidly than in other places. Therefore, that potion of
the graph may show us numbers that do not follow the common rule of higher
digits having higher probabilities.
Fig 29. Height growth of boys from birth to 18.5 years of age
Let us fix the problem mentioned above. This needs to be done because
equal x-value intervals are one of the conditions of calculating the probabili-
ties correctly. We will use the graph above for this purpose. In particular, we
will complete the table above by inserting additional values into it in such a
way, that each pair of consecutive x-values will be exactly one unit apart. The
corresponding y-values will be approximated by using the graph.
As we can see, all the horizontal colored lines cross the graph at points that
have particular x-values. To be more exact, the x-values of those points are 5
units apart from each other. Having 222 x-value units gives us approximately
222 ÷ 5 ≈ 44 horizontal lines. These lines show the corresponding y-values with
relatively high accuracy. The rest of the values can be approximated as well.
Table 12 shows the modified values. The first column contains x-values, which
are exactly one unit apart from each other. The original y-values are unchanged
and they are in the second column now. All of them have two decimal places.
The approximated y-values are whole numbers in the second column as well.
The last column shows the first digit of the corresponding y-value from the
second column.
We will exclude the point with coordinates (0, 0) from our table, because it
does not belong to our domain. We also will exclude points that have x-values
of 218 and higher because their corresponding y-values exceed our y-value limit
of 1000.
Table 12
Modified values from the above table
x- y- first x- y- first x- y- first
values values digit values values digit values values digit
1 5.12 5 51 406 4 101 614 6
2 33.12 3 52 412 4 102 617.76 6
3 58.64 5 53 415 4 103 621 6
4 85.52 8 54 420.81 4 104 625 6
5 102.01 1 55 427 4 105 629 6
6 122.24 1 56 429 4 106 633 6
7 136.72 1 57 432 4 107 637 6
8 150.16 1 58 436 4 108 641.44 6
9 162.32 1 59 440 4 109 645 6
10 168.96 1 60 445.92 4 110 649 6
11 174.24 1 61 450 4 111 654 6
12 188.41 1 62 455 4 112 658 6
13 200 2 63 461 4 113 662 6
14 204 2 64 465 4 114 666.48 6
15 211.28 2 65 471 4 115 669 6
16 221 2 66 476.32 4 116 672 6
17 225 2 67 482 4 117 675 6
18 233.92 2 68 486 4 118 678 6
19 244 2 69 488 4 119 681 6
20 250 2 70 493 4 120 684.08 6
21 258.81 2 71 495 4 121 689 6
22 269 2 72 499.92 4 122 693 6
23 275 2 73 503 5 123 697 6
24 279.44 2 74 508 5 124 700 7
25 286 2 75 512 5 125 703 7
26 290 2 76 516 5 126 705.52 7
27 292 2 77 521 5 127 709 7
28 294 2 78 526.48 5 128 713 7
29 301 3 79 530 5 129 716 7
30 302.24 3 80 534 5 130 720 7
31 309 3 81 539 5 131 724 7
32 316 3 82 542 5 132 727.36 7
33 320 3 83 546 5 133 730 7
34 328 3 84 550.32 5 134 733 7
35 334 3 85 554 5 135 737 7
36 336.56 3 86 558 5 136 740 7
37 343 3 87 563 5 137 743 7
38 349 3 88 567 5 138 746 7
39 351 3 89 572 5 139 750 7
40 355 3 90 576.56 5 140 754 7
41 360 3 91 581 5 141 759 7
42 363.36 3 92 585 5 142 763 7
43 368 3 93 589 5 143 767 7
44 374 3 94 594 5 144 772 7
45 379 3 95 598 5 145 777 7
46 383 3 96 601.68 6 146 782 7
47 389 3 97 604 6 147 787 7
48 392.41 3 98 606 6 148 793 7
49 398 3 99 608 6 149 798 7
50 400 4 100 611 6 150 800.24 8
x- y- first x- y- first x- y- first

151 805 8 174 917.84 9 197 974 9
152 809 8 175 921 9 198 975.28 9
153 814 8 176 924 9 199 978 9
154 818 8 177 927 9 200 981 9
155 823 8 178 930 9 201 984 9
156 827.12 8 179 933 9 202 987 9
157 832 8 180 936.16 9 203 990 9
158 837 8 181 939 9 204 993.84 9
159 842 8 182 942 9 205 994 9
160 846 8 183 946 9 206 994 9
161 851 8 184 948 9 207 994 9
162 856.72 8 185 950 9 208 995 9
163 859 8 186 953.68 9 209 995 9
164 862 8 187 955 9 210 995.04 9
165 866 8 188 957 9 211 996 9
166 869 8 189 960 9 212 996 9
167 873 8 190 962 9 213 996 9
168 876.64 8 191 965 9 214 997 9
169 887 8 192 967.92 9 215 997 9
170 893 8 193 969 9 216 997.61 9
171 898 8 194 970 9 217 999 9
172 906 9 195 972 9
173 910 9 196 973 9
Now, all the information needed for the final calculation is presented. Let us
summarize everything we have so far using Table 13.
Table 13
Analysis of y−values’ first digits
Digit Count Pk from the graph Pk values
√
(Count / Sum) for y = x
1 8 0.03686636 0.03030303
2 16 0.07373272 0.05050505
3 22 0.10138249 0.07070707
4 23 0.10599078 0.09090909
5 25 0.11520737 0.11111111
6 28 0.12903226 0.13131313
7 26 0.11981567 0.15151515
8 23 0.10599078 0.17171717
9 46 0.21198157 0.19191919
Sum 217 1.00000000 1.00000000
As we can see, some corresponding Pk values are closer to each other than
others. The biggest differences occur between probabilities of the digits 7 and 8.
As it was mentioned above, the top portion of the graph has a couple of places
where the slope changes faster than in other places. Thus, the probabilities of
the digits 7 and 8 do not follow the common rule of higher digits having higher
probabilities. Therefore, their differences from the corresponding probabilities
of the square root function are bigger than those of other digits.
√
Fig 30. Young males hight increase and the function y = 65.32 x
With a couple of exceptions, the graph reflecting the height increase among
√
young males is very similar to the specific part of the function y = 65.32 x.
Both graphs are shown in Figure 30.
2.5. Pk of the logarithmic function y = loga x
The function y = loga x is the most challenging one in this chapter.
Fig 31. Function y = log2 x

We will talk about this function having base a as a positive number greater
than 1. As it was done before with other functions, we will look at the function
y = loga x on the range [1, ∞). In addition, we will pick an interval [a, ∞) for
the domain, so they both will match each other.
To be able to graph the function, we have to assign a particular value to a.
First, let a = 2; then we will assign a = 3.
Due to that fact that the logarithmic function with any value of a grows
much slower than all previous ones that we discussed before, we will graph both
functions y = log2 x and y = log3 x on the limited range [1, 10). However, the
range [1, ∞) will be used for calculating the Pk values initially.
The graph of the function y = log2 x is shown in Figure 31. It is obvious that
bigger digits have higher probabilities. We will do more work later to prove it.
Let us look at the function y = log3 x. Figure 32 shows the function and yet
again, bigger digits have higher probabilities.
Fig 32. Function y = log3 x
Thus, let us prove it. Let f (x) = loga x. According to the definition of the
logarithmic function, f −1 (x) = ax . Next,
n−1 n−1 n−1 n−1
a(k+1)·10 − ak·10 (a10 )k+1 − (a10 )k
Pk = n−1 = ;
a10·10 − a10n−1 (a10n−1 )10 − a10n−1
n−1
Let us use a substitution c = a10 . Then,
n−1 n−1
(a10 )k+1 − (a10 )k ck+1 − ck ck (c − 1)
n−1 n−1 = = ;
(a10 )10 − a10 c10 − c c(c9 − 1)
Next, let us modify the denominator.
c9 − 1 = (c3 )3 − (13 )3 = (c3 − 13 )(c6 + c3 + 1) = (c − 1)(c2 + c + 1)(c6 + c3 + 1)
= (c − 1)(c8 + c7 + c6 + c5 + c4 + c3 + c2 + c + 1);
Then, we will bring it back.
ck (c − 1) ck (c − 1)
9
=
c(c − 1) c(c − 1)(c + c + c + c5 + c4 + c3 + c2 + c + 1)
8 7 6
ck
=
c(c8 + c7 + c6 + c5 + c4 + c3 + c2 + c + 1)
ck ck
= = 9 ;
c9 + c8 + c7 + c6 + c5 + c4 + c3 + c2 + c
P i
i=1 c
Let us bring our variables a and n back.
n−1 n−1
ck (a10 )k ak·10
P9 = P9 = P9 ;
i=1 ci i=1 (a10n−1 )i i=1 ai·10n−1
Finaly,
n−1
ak·10
Pk = P9 (11)
i·10n−1
i=1 a
Let us analyze the result. The numerator of the fraction is one of the sum’s
components, located in the fraction’s denominator. Thus, the fractions value is
always less than 1. Let us check if the sum of all the Pk values equals to 1.
9 9 n−1 P9 n−1
X X ak·10 ak·10
Pk = P9 = Pk=1
9 = 1;
k=1 k=1 i=1 ai·10n−1 i=1 ai·10n−1
Since the top and the bottom sums are equal to each other, the fraction
equals to 1.
Let us look at the Formula 11. It contains the exponent n, which is the
reference to the function’s range. It means that the function’s f dp depends on
the range used for calculations. In addition, the formula contains the base a.
Thus, we can conclude that the f dp of the logarithmic function depends on its
base.
Let us start from figuring out what is going on in the range [1, 10) first. It
means that we will assign n = 1 for now. In addition, since the variable a is a
part of the formula, we will first assign a = 2, and then a = 3.
If n = 1, then
1−1 0
ak·10 ak·10 ak
Pk = P9 = 9 = 9 ;
i·101−1 i·100
P P i
i=1 a i=1 a i=1 a
If a = 2, then
ak 2k 2k 2k−1
Pk = P9 = P9 = = ;
i=1 ai i=1 2i 1022 511
If a = 3, then
ak 3k 3k 3k−1
Pk = P9 = P9 = = ;
i=1 ai i=1 3i 29523 9841
Let us look at a table with Pk values for the functions y = log2 x and y = log3 x
on the range [1, 10). Table 14 shows the data.
Table 14
Pk values for y = log2 x and y = log3 x on the range [1, 10).
Digit Pk for y = log2 x Pk for y = log3 x

1 0.00195695 0.00010162
2 0.00391389 0.00030485
3 0.00782779 0.00091454
4 0.01565558 0.00274362
5 0.03131115 0.00823087
6 0.06262231 0.02469261
7 0.12524462 0.07407784
8 0.25048924 0.22223351
9 0.50097847 0.66670054
Sum 1.00000000 1.00000000
While looking at the Pk formulas and graphs above, we could already see
that bigger digits have higher probabilities. This conclusion is proven again by
our table for n = 1. In addition, it looks like it will be true for higher n-values
as well.
As we can see, each Pk value is bigger than the previous one exactly a times.
Moreover, the difference between first and last f dp values for both a is much
higher than in all previous cases of this work. Furthermore, it looks like this
difference is even higher for functions with higher bases.
After the formulas above were defined, let us see graphs of the functions
k−1
for a = 2 and a = 3 on the range [1, 10). The functions are Pk = 2511 and
k−1
Pk = 39841 . Both graphs are shown in Figure 33.
2k−1 3k−1
Fig 33. Functions Pk = 511
and Pk = 9841
Let us see now how the formulas will look like on the range [10, 100). Thus,
we will assign n = 2. Then,
2−1 1
ak·10 ak·10 a10k
Pk = P9 2−1
= P9 1
= P9 ;
i·10 i·10 10i
i=1 a i=1 a i=1 a
If a = 2, then
a10k 210k (210 )k 1024k

Pk = P9 = P9 = P9 = P9 ;
10i 10i 10 i i
i=1 a i=1 2 i=1 (2 ) i=1 1024
If a = 3, then
a10k 310k (310 )k 59049k

Pk = P 9 = P9 = P9 = P9 ;
10i 10i 10 i i
i=1 a i=1 3 i=1 (3 ) i=1 59049
We should now look at a table with Pk values for the functions y = log2 x
and y = log3 x on the range [10, 100). Table 15 shows the data.
As we can see, both P9 values almost equal to 1. The rest of Pk values are
very small.
Table 15
Pk values for y = log2 x and y = log3 x on the range [10, 100).
Digit Pk for y = log2 x Pk for y = log3 x

1 8.263728 · 1025 6.765381 · 1039
2 8.462058 · 1022 3.994890 · 1034
3 8.665147 · 1019 2.358943 · 1029
4 8.873111 · 1016 1.392932 · 1024
5 9.086065 · 1013 8.225124 · 1020
6 9.304131 · 1010 4.856853 · 1015
7 9.527430 · 107 2.867923 · 1010
8 9.756088 · 104 1.693480 · 105
9 9.990234 · 101 9.999831 · 101
Sum 1.00000000 1.00000000
k k
Let us also compare the functions Pk = P91024 and Pk = P959049
1024i 59049i
i=1 i=1
while looking at their graphs. Both graphs corresponding to the range [10, 100)
are presented in Figure 34. As we can see, they are located really close to each
other.
k k
Fig 34. Functions Pk = P91024 and Pk = P959049
1024i 59049i
i=1 i=1
However, the curve connecting the points of the Pk values of y = log3 x func-
tion is located at the right of the curve connecting the points of the Pk values
of y = log2 x function in between k values of 8 and 9. It is the only way to
differentiate them in the graph. As it was mentioned above, all the Pk values
except P9 of both functions are so small, that they almost equal to zero.
Let us think about a proper example that can be used in this section in such a
way that it describes a process that has logarithmic growth and has Pk values
equal or approximately equal to ours.
First, let us analyze what will happen if we stretch or squeeze our initial
y = loga x function. In particular, we need to know if it affects its Pk values.
Only if the Pk values stay unchanged, we can try to find a proper example.
To make our steps a little bit easier later, let us divide the whole function by
a constant h instead of multiplying by it as we did before. Depending on value
of h, we either stretch or squeeze our function.
Let y = logha x . Then, hy = loga x. Let us switch x and y: hx = loga y. Then,
y = ahx . Let f (x)−1 = ahx . Then,
n−1 n−1 n−1 n−1
ah(k+1)·10 − ah·k·10 (ah )(k+1)·10 − (ah )k·10
Pk = n−1 = ;
ah·10·10 − ah·10n−1 (ah )10·10n−1
− (ah )10n−1
Let us use a substitution w = ah . Since a and h are constants, w is a constant
as well. Then:
n−1 n−1
w(k+1)·10 − wk·10
Pk = n−1 ;
w 10·10 − w10n−1
As we can see, we came to the very first step of getting the Pk values for a
logarithmic function. The only difference is that the formula has a base w now,
not a base a as it was before.
We know that logarithmic functions with different bases have different Pk
values. In other words, stretching or squeezing of the logarithmic function will
change its Pk values, which tells us that we will have to change the base of our
log function in a certain way if h 6= 1. In particular, the base a will be changed
for the base ah .
Next, let us check what will happen if we multiply or divide the variable x
x
by a constant directly. Let us have y = loga ( m ), where m is a constant. Then
we have:
x
y = loga ( ) = loga x − loga m;
m
Since both a and m are constants, loga m is a constant as well. As we can see,
dividing or multiplying the variable x by a constant will result in a vertical shift
of the initial function. Let us assign q = loga m and check how a vertical shift
effects the Pk values.
Let y = loga x − q. Then y + q = loga x. After we switch x and y, we get
loga y = x + q. Then, y = ax+q . Let f (x)−1 = ax+q . Then,
n−1 n−1 n−1 n−1

a(k+1)·10 +q − ak·10 +q a(k+1)·10 · aq − ak·10 · aq
Pk = =
a10·10n−1 +q − a10n−1 +q a10·10n−1 · aq − a10n−1 · aq
n−1 n−1 n−1 n−1
aq (a(k+1)·10 − ak·10 ) a(k+1)·10 − ak·10
= = ;
aq (a10·10n−1 − a10n−1 ) a10·10n−1 − a10n−1
Obviously, we came back to the first step of getting the Pk values for our ini-
tial function. The last fraction above is identical to the one with no vertical shift.
It means that a vertical shift of a logarithmic function does not change the func-
tion’s initial Pk values. Thus, we do not have to do anything about multiplying
or dividing the variable x by a constant if we face a case like that in our example.
It looks like we have enough information by now to start thinking about our
example. We had a chance to talk about bacterial growth before. Let us talk
about population growth now[5].
We will be using the formula p = 1600e0.025T , where p is the number of
people at time T , 1600 is the number of people when T = 0, and 0.025 is the
population growth rate. Let us isolate the variable T . It will allow us to get a
logarithmic function.
p p 1 p p
= e0.025T ; 0.025T = ln( ); T = ln( ); T = 40 ln( );
1600 1600 40 1600 1600
p
Thus, T (p) = 40 ln( 1600 ) is the function that we will be using in our example.
This equation shows us how many years are needed to get a certain population.
Let us recall that a log function changes its Pk values depending on its range.
To make our work meaningful, let us choose the range [1, 10). All other ranges
are much harder to analyze.
Let us calculate the initial and final values of p in such a way that they will
be starting and ending points of our domain.
First, let T = 1. Then,
p 1 p p p
1 = 40 ln( ); = ln( ); ln( ) = 0.025; = e0.025 ;
1600 40 1600 1600 1600
p = 1600e0.025 ≈ 1600 · 1.0253 = 1640.48;

We will approximate the initial p value to 1640, so it would bring us to the
initial T value that is a little bit less than 1.
Next, let T = 10. Then,
p 1 p p p
10 = 40 ln( ); = ln( ); ln( ) = 0.25; = e0.25 ;
1600 4 1600 1600 1600
p = 1600e0.25 ≈ 1600 · 1.2840 = 2054.4;

We will approximate the final p value to 2055, so it would give us the final T
value that is a little bit greater than 10.
The difference 2055−1640 = 415 will give us enough steps for our calculations.
We will calculate a new T value as soon as the number of people gets bigger by
one. If calculated T values exceed our range [1, 10), we will not use them for our
f dp calculations. Table 16 shows the results.
Table 16
Calculated T (p) values and their first digits
p- T- first p- T- first p- T- first

1640 0.9877 1690 2.1890 2 1740 3.3553 3
1641 1.0121 1 1691 2.2127 2 1741 3.3782 3
1642 1.0365 1 1692 2.2363 2 1742 3.4012 3
1643 1.0608 1 1693 2.2599 2 1743 3.4242 3
1644 1.0851 1 1694 2.2836 2 1744 3.4471 3
1645 1.1095 1 1695 2.3072 2 1745 3.4700 3
1646 1.1338 1 1696 2.3308 2 1746 3.4930 3
1647 1.1581 1 1697 2.3543 2 1747 3.5159 3
1648 1.1824 1 1698 2.3779 2 1748 3.5387 3
1649 1.2066 1 1699 2.4014 2 1749 3.5616 3
1650 1.2309 1 1700 2.4250 2 1750 3.5845 3
1651 1.2551 1 1701 2.4485 2 1751 3.6073 3
1652 1.2793 1 1702 2.4720 2 1752 3.6302 3
1653 1.3035 1 1703 2.4955 2 1753 3.6530 3
1654 1.3277 1 1704 2.5190 2 1754 3.6758 3
1655 1.3519 1 1705 2.5425 2 1755 3.6986 3
1656 1.3761 1 1706 2.5659 2 1756 3.7214 3
1657 1.4002 1 1707 2.5894 2 1757 3.7442 3
1658 1.4243 1 1708 2.6128 2 1758 3.7669 3
1659 1.4485 1 1709 2.6362 2 1759 3.7897 3
1660 1.4726 1 1710 2.6596 2 1760 3.8124 3
1661 1.4966 1 1711 2.6830 2 1761 3.8351 3
1662 1.5207 1 1712 2.7063 2 1762 3.8578 3
1663 1.5448 1 1713 2.7297 2 1763 3.8805 3
1664 1.5688 1 1714 2.7530 2 1764 3.9032 3
1665 1.5929 1 1715 2.7764 2 1765 3.9259 3
1666 1.6169 1 1716 2.7997 2 1766 3.9485 3
1667 1.6409 1 1717 2.8230 2 1767 3.9712 3
1668 1.6649 1 1718 2.8463 2 1768 3.9938 3
1669 1.6888 1 1719 2.8696 2 1769 4.0164 4
1670 1.7128 1 1720 2.8928 2 1770 4.0390 4
1671 1.7367 1 1721 2.9161 2 1771 4.0616 4
1672 1.7607 1 1722 2.9393 2 1772 4.0842 4
1673 1.7846 1 1723 2.9625 2 1773 4.1068 4
1674 1.8085 1 1724 2.9857 2 1774 4.1293 4
1675 1.8324 1 1725 3.0089 3 1775 4.1519 4
1676 1.8563 1 1726 3.0321 3 1776 4.1744 4
1677 1.8801 1 1727 3.0553 3 1777 4.1969 4
1678 1.9040 1 1728 3.0784 3 1778 4.2194 4
1679 1.9278 1 1729 3.1016 3 1779 4.2419 4
1680 1.9516 1 1730 3.1247 3 1780 4.2644 4
1681 1.9754 1 1731 3.1478 3 1781 4.2869 4
1682 1.9992 1 1732 3.1709 3 1782 4.3093 4
1683 2.0230 2 1733 3.1940 3 1783 4.3317 4
1684 2.0467 2 1734 3.2171 3 1784 4.3542 4
1685 2.0705 2 1735 3.2402 3 1785 4.3766 4
1686 2.0942 2 1736 3.2632 3 1786 4.3990 4
1687 2.1179 2 1737 3.2862 3 1787 4.4214 4
1688 2.1416 2 1738 3.3093 3 1788 4.4438 4
1689 2.1653 2 1739 3.3323 3 1789 4.4661 4

1790 4.4885 4 1840 5.5905 5 1890 6.6629 6
1791 4.5108 4 1841 5.6122 5 1891 6.6841 6
1792 4.5331 4 1842 5.6339 5 1892 6.7052 6
1793 4.5555 4 1843 5.6556 5 1893 6.7264 6
1794 4.5778 4 1844 5.6773 5 1894 6.7475 6
1795 4.6001 4 1845 5.6990 5 1895 6.7686 6
1796 4.6223 4 1846 5.7207 5 1896 6.7897 6
1797 4.6446 4 1847 5.7424 5 1897 6.8108 6
1798 4.6669 4 1848 5.7640 5 1898 6.8319 6
1799 4.6891 4 1849 5.7857 5 1899 6.8530 6
1800 4.7113 4 1850 5.8073 5 1900 6.8740 6
1801 4.7335 4 1851 5.8289 5 1901 6.8951 6
1802 4.7557 4 1852 5.8505 5 1902 6.9161 6
1803 4.7779 4 1853 5.8721 5 1903 6.9371 6
1804 4.8001 4 1854 5.8937 5 1904 6.9581 6
1805 4.8223 4 1855 5.9152 5 1905 6.9791 6
1806 4.8444 4 1856 5.9368 5 1906 7.0001 7
1807 4.8666 4 1857 5.9583 5 1907 7.0211 7
1808 4.8887 4 1858 5.9799 5 1908 7.0421 7
1809 4.9108 4 1859 6.0014 6 1909 7.0630 7
1810 4.9329 4 1860 6.0229 6 1910 7.0840 7
1811 4.9550 4 1861 6.0444 6 1911 7.1049 7
1812 4.9771 4 1862 6.0659 6 1912 7.1258 7
1813 4.9992 4 1863 6.0874 6 1913 7.1468 7
1814 5.0212 5 1864 6.1088 6 1914 7.1677 7
1815 5.0433 5 1865 6.1303 6 1915 7.1886 7
1816 5.0653 5 1866 6.1517 6 1916 7.2094 7
1817 5.0873 5 1867 6.1732 6 1917 7.2303 7
1818 5.1093 5 1868 6.1946 6 1918 7.2512 7
1819 5.1313 5 1869 6.2160 6 1919 7.2720 7
1820 5.1533 5 1870 6.2374 6 1920 7.2929 7
1821 5.1753 5 1871 6.2588 6 1921 7.3137 7
1822 5.1972 5 1872 6.2801 6 1922 7.3345 7
1823 5.2192 5 1873 6.3015 6 1923 7.3553 7
1824 5.2411 5 1874 6.3229 6 1924 7.3761 7
1825 5.2631 5 1875 6.3442 6 1925 7.3969 7
1826 5.2850 5 1876 6.3655 6 1926 7.4177 7
1827 5.3069 5 1877 6.3868 6 1927 7.4384 7
1828 5.3288 5 1878 6.4082 6 1928 7.4592 7
1829 5.3506 5 1879 6.4294 6 1929 7.4799 7
1830 5.3725 5 1880 6.4507 6 1930 7.5007 7
1831 5.3943 5 1881 6.4720 6 1931 7.5214 7
1832 5.4162 5 1882 6.4933 6 1932 7.5421 7
1833 5.4380 5 1883 6.5145 6 1933 7.5628 7
1834 5.4598 5 1884 6.5357 6 1934 7.5835 7
1835 5.4816 5 1885 6.5570 6 1935 7.6041 7
1836 5.5034 5 1886 6.5782 6 1936 7.6248 7
1837 5.5252 5 1887 6.5994 6 1937 7.6455 7
1838 5.5470 5 1888 6.6206 6 1938 7.6661 7
1839 5.5687 5 1889 6.6418 6 1939 7.6867 7

1940 7.7074 7 1979 8.5035 8 2018 9.2841 9
1941 7.7280 7 1980 8.5237 8 2019 9.3039 9
1942 7.7486 7 1981 8.5439 8 2020 9.3238 9
1943 7.7692 7 1982 8.5641 8 2021 9.3436 9
1944 7.7898 7 1983 8.5843 8 2022 9.3633 9
1945 7.8103 7 1984 8.6045 8 2023 9.3831 9
1946 7.8309 7 1985 8.6246 8 2024 9.4029 9
1947 7.8514 7 1986 8.6448 8 2025 9.4226 9
1948 7.8720 7 1987 8.6649 8 2026 9.4424 9
1949 7.8925 7 1988 8.6850 8 2027 9.4621 9
1950 7.9130 7 1989 8.7051 8 2028 9.4819 9
1951 7.9335 7 1990 8.7252 8 2029 9.5016 9
1952 7.9540 7 1991 8.7453 8 2030 9.5213 9
1953 7.9745 7 1992 8.7654 8 2031 9.5410 9
1954 7.9950 7 1993 8.7855 8 2032 9.5607 9
1955 8.0155 8 1994 8.8056 8 2033 9.5804 9
1956 8.0359 8 1995 8.8256 8 2034 9.6000 9
1957 8.0564 8 1996 8.8457 8 2035 9.6197 9
1958 8.0768 8 1997 8.8657 8 2036 9.6393 9
1959 8.0972 8 1998 8.8857 8 2037 9.6590 9
1960 8.1176 8 1999 8.9057 8 2038 9.6786 9
1961 8.1380 8 2000 8.9257 8 2039 9.6982 9
1962 8.1584 8 2001 8.9457 8 2040 9.7178 9
1963 8.1788 8 2002 8.9657 8 2041 9.7375 9
1964 8.1992 8 2003 8.9857 8 2042 9.7570 9
1965 8.2195 8 2004 9.0057 9 2043 9.7766 9
1966 8.2399 8 2005 9.0256 9 2044 9.7962 9
1967 8.2602 8 2006 9.0456 9 2045 9.8158 9
1968 8.2806 8 2007 9.0655 9 2046 9.8353 9
1969 8.3009 8 2008 9.0854 9 2047 9.8549 9
1970 8.3212 8 2009 9.1053 9 2048 9.8744 9
1971 8.3415 8 2010 9.1252 9 2049 9.8939 9
1972 8.3618 8 2011 9.1451 9 2050 9.9134 9
1973 8.3821 8 2012 9.1650 9 2051 9.9330 9
1974 8.4023 8 2013 9.1849 9 2052 9.9525 9
1975 8.4226 8 2014 9.2048 9 2053 9.9719 9
1976 8.4428 8 2015 9.2246 9 2054 9.9914 9
1977 8.4631 8 2016 9.2445 9 2055 10.0109
1978 8.4833 8 2017 9.2643 9
Before we summarize the results from the table above, let us prepare the data
that will be compared to the mentioned above results. We already discovered
that the Pk values of a function y = h1 loga ( m
x
) are the same as the Pk values
of a function y = log(ah ) (x) and the constant m has no effect on it.
In the example above, a = e because we are dealing with natural log there
and h = 0.025. Thus, ah = e0.025 ≈ 1.0253. It means that we will be calculating
the Pk values of the function y = log1.0253 x. Moreover, we are expecting that
the values will be at least approximately equal to those from the table above.
Both groups of the Pk values are presented in Table 17. The middle column
x
shows the result of the calculations based on the formula y = 40 ln( 1600 ) and
the table above. The last column shows the results based on the general Pk
formula that was defined for any log function and the formula y = log1.0253 x.
The range [1, 10) was used in both cases.
Table 17
Analysis of T (p)−values’ first digits
x
Digit Count Pk for y = 40 ln( 1600 ) Pk for y = log1.0253 x
(Count / Sum)
1 42 0.10144928 0.10033451
2 42 0.10144928 0.10287297
3 44 0.10628019 0.10547566
4 45 0.10869565 0.10814419
5 45 0.10869565 0.11088024
6 47 0.11352657 0.11368551
7 49 0.11835749 0.11656175
8 49 0.11835749 0.11951077
9 51 0.12318841 0.12253439
Sum 217 1.00000000 1.00000000
As we can see, the corresponding Pk values are really close to each other. The
last column shows numbers that are more accurate because their calculations
were derived directly from the formula. The middle column has numbers that
are less accurate because the log function used in the example was approximated
to the discrete one with steps of one unit. Having more steps would definitely
help to get numbers that are more accurate.
We also can see that the difference between the first and last Pk values is not
as big as in all our previous calculations. We already discovered that a bigger a
value causes bigger gaps between P1 and P9 . However, the base of the function
log1.0253 x is very small. It is the reason why the differences between Pk values
in the table above are not that big.
To finish our comparison, let us graph both functions. Unfortunately, we will
not be able to graph them in the same xy-plane without doing extra tricks
due to a big gap in between their domains that we have to pick. The function
x
y = 40 ln( 1600 ) is presented above on the domain [1640, 2055].
Let us calculate the domain of the function y = log1.0253 x in such a way that
it will match to the range [1, 10).
1.02531 = 1.0253; 1.025310 ≈ 1.2838;
Thus, the range is [1.0253, 1.2838). If we skip certain portions of the x-axis,
we will be able to locate both graphs together.
We are ready to look at the graphs now. Both of them are shown in Figure 35.
Please, notice that their actual domains are really far from each other. As we
can see, both log functions are stretched so much that their graphed portions
look almost like straight lines. That is why their Pk values are close to those of
straight lines.
x
Fig 35. Functions y = log1.0253 x and y = 40 ln( 1600 )
a
2.6. Pk of the reciprocal function y = x
The reciprocal function y = xa , where a is a constant, is the last one in this

chapter. Since we are graphing all our functions in the first quadrant only, the
constant a for this function will be always positive.
We will consider three cases: a > 1, a = 1, and 0 < a < 1. We also should keep
in mind that our range is always restricted to [1, ∞), that is why our domain
for this function is the interval (0, a].
All three functions are graphed on the range [1, 10). Let us look at each case
separately.
The function y = 25 1
x is shown in Figure 36, the function y = x is shown in
0.25
Figure 37, and the function y = x is shown in Figure 38.
As we can see, two of the graphs are stretched horizontally. Thus, all three
functions look alike now. We can also see that in all three cases smaller digits
have higher f dp values. We already know that this feature depends on the
function’s shape. Let us calculate the values.
25
Fig 36. Function y = x
1
0.25
a
The general formula 1 will be used. First, let y = x. Then, if we switch x
and y, x = ay . Thus, y = xa again and f −1 (x) = xa .
a a ak a(k+1) ak−a(k+1)
(k+1)·10n−1 − k·10n−1 k(k+1)·10n−1 − k(k+1)·10n−1 k(k+1)·10n−1
Pk = a a = a 10a = a−10a
10·10n−1 − 10n−1 10·10n−1 − 10·10n−1 10·10n−1
ak−ak−a −a
k(k+1)·10n−1 k(k+1)·10n−1 −10 · 10n−1 a 10
= −9a = −9a = n−1
= ;
10·10n−1 10·10n−1
−9ak(k + 1) · 10 9k(k + 1)
Thus,
10
Pk = (12)
9k(k + 1)
We can see that the Pk values of the function will be the same on any range
and the constant a does not change them as well. The graph of the Pk function
is shown in Figure 39.
It is obvious now that smaller first digits have higher probabilities. Besides,
all the probabilities are positive numbers less than 1. Let us check whether the
sum of all nine probabilities totals to 1. First, let us modify the formula.
10
Fig 39. Function y = 9k(k+1)
1 A B A(k + 1) Bk
= + = +
k(k + 1) k k+1 k(k + 1) k(k + 1)
A(k + 1) + Bk Ak + A + Bk k(A + B) + A
= = = ;
k(k + 1) k(k + 1) k(k + 1)
1
Since k(k+1) = k(A+B)+A
k(k+1) , k(A + B) + A = 1. Thus, A + B = 0 and A = 1.
1
Consequently, B = −1 and k(k+1) = k1 − (k+1)
1
.
Thus,
10 10 1 10 1 1
Pk = = · = ( − );
9k(k + 1) 9 k(k + 1) 9 k k+1
Then,
9 9 9
X X 10 1 1 10 X 1 1 10 1 1 1 1
Pk = ( − )= ( − )= ( − + −
9 k k+1 9 k k+1 9 1 1+1 2 2+1
k=1 k=1 k=1
1 1 1 1 1 1 1 1 1 1 1 1 1 1
+ − + − + − + − + − + − + − )
3 3+1 4 4+1 5 5+1 6 6+1 7 7+1 8 8+1 9 9+1
10 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
= (1 − + − + − + − + − + − + − + − + − )
9 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 10
10 1 10 9
= (1 − ) = · = 1;
9 10 9 10
P9
Thus, k=1 Pk = 1, so we have obtained one of the proofs.
Let us do one more proof that will be done by graphing after the following
step is performed.
10 1 1 10 10
Pk = ( − )= − ;
9 k k+1 9k 9(k + 1)
10
Then, we will graph the function f (x) = 9x on the domain from 1 to 10. Fig-
ure 40 shows the graph.
10
Fig 40. The function f (x) = 9x
10 10
We will use the graph for locating the values 9k and 9(k+1) where k is a whole
number from 1 to 9. Both values can be found on the y-axis. After picking a
10 10
particular k, subtracting 9(k+1) from 9k will give us the corresponding Pk value.
10 10
The graph shows the differences between the values 9k and 9(k+1) for all nine
values of k. All the differences are equal to the corresponding Pk values. They
are labeled with the corresponding colors that are taken from the ruler described
above.
The graph clearly shows that the sum of all nine Pk values equals to the
difference between f (1) and f (10). Thus,
10 10 10 10 100 10 90
f (1) − f (10) = − = − = − = = 1;
9 · 1 9 · 10 9 90 90 90 90
P9
It is another proof of the statement k=1 Pk = 1.
Let us calculate the Pk values of our function and list them in Table 18. We
should recall that the values do not depend on the constant a and on the chosen
range. It means that numbers presented in the table will stay the same as long
as we are dealing with any reciprocal function y = xa .
Table 18
Pk values of the reciprocal function
a
Digit Pk for y = x
1 0.55555556
2 0.18518519
3 0.09259259
4 0.05555556
5 0.03703704
6 0.02645503
7 0.01984127
8 0.01543210
9 0.01234568
Sum 1.00000000
Before we talk about how the reciprocal function can be used to model a real
life situation, we should check how a horizontal shift would affect the Pk values
of the function.
Our intuition should tell us that a horizontal shift of any function should not
change its Pk values because a horizontally shifted function will be reflected on
the y-axis in exactly the same way as it was reflected initially. However, we just
like proofs.
a
Let us consider a case y = x−h where h is a constant which shows a horizontal
shift.
a
First, we will switch x and y. It will give us x = y−h . Then, y − h = xa and
y = xa + h. Thus, f −1 (x) = xa + h.
f −1 ((k + 1) · 10n−1 ) − f −1 (k · 10n−1 )

Pk =
f −1 (10n ) − f −1 (10n−1 )
a a a a
( (k+1)·10n−1 + h) − ( k·10n−1 + h) (k+1)·10n−1 + h − k·10n−1 −h
= =
( 10an a
+ h) − ( 10n−1 + h) a a
10n + h − 10n−1 − h
a a
(k+1)·10n−1 − k·10n−1
= a a ;
10n − 10n−1
Thus, we came back to the initial Pk formula. It means that the Pk values of
our function will not be affected by a horizontal shift.
Let us talk about the actual application now. We will be solving a problem
525
about a scuba diver [2]. The formula t = d−10 shows the greatest amount of
time in minutes that a scuba diver can take to rise toward the water surface
without stopping for decompression, where d is the depth, in meters, of the
diver.
As we can see, it is a reciprocal function with a horizontal shift. In addition,
since the formula represents a real-life process, its t-value cannot be negative.
Moreover, the denominator cannot be equal to zero. It means that the domain
should be initially restricted to (10, ∞).
Let us calculate the f dp of the t-values of the function on the range [1, 100).
We also should find the domain that will match to this range.
525
1= ; d − 10 = 525; d = 535;
d − 10
525 525
100 = ; d − 10 = = 5.25; d = 15.25;
d − 10 100
We will pick a domain interval [15, 535]. However, if some calculated t-values
exceed the range [1, 100), we will not include them in our further calculations.
We also should check that selected by us domain is located inside of the initial
domain (10, ∞), which is the case.
We will plug every integer from 15 to 535 into the variable d in the formula
525
t(d) = d−10 , calculate the corresponding t-value, retrieve the t-value’s first digit,
and list all the results in Table 19.
Table 19
Calculated t(d) values and their first digits
d- t- first d- t- first d- t- first

15 105 39 18.10345 1 63 9.90566 9
16 87.5 8 40 17.5 1 64 9.722222 9
17 75 7 41 16.93548 1 65 9.545455 9
18 65.625 6 42 16.40625 1 66 9.375 9
19 58.33333 5 43 15.90909 1 67 9.210526 9
20 52.5 5 44 15.44118 1 68 9.051724 9
21 47.72727 4 45 15 1 69 8.898305 8
22 43.75 4 46 14.58333 1 70 8.75 8
23 40.38462 4 47 14.18919 1 71 8.606557 8
24 37.5 3 48 13.81579 1 72 8.467742 8
25 35 3 49 13.46154 1 73 8.333333 8
26 32.8125 3 50 13.125 1 74 8.203125 8
27 30.88235 3 51 12.80488 1 75 8.076923 8
28 29.16667 2 52 12.5 1 76 7.954545 7
29 27.63158 2 53 12.2093 1 77 7.835821 7
30 26.25 2 54 11.93182 1 78 7.720588 7
31 25 2 55 11.66667 1 79 7.608696 7
32 23.86364 2 56 11.41304 1 80 7.5 7
33 22.82609 2 57 11.17021 1 81 7.394366 7
34 21.875 2 58 10.9375 1 82 7.291667 7
35 21 2 59 10.71429 1 83 7.191781 7
36 20.19231 2 60 10.5 1 84 7.094595 7
37 19.44444 1 61 10.29412 1 85 7 7
38 18.75 1 62 10.09615 1 86 6.907895 6

87 6.818182 6 137 4.133858 4 187 2.966102 2
88 6.730769 6 138 4.101563 4 188 2.949438 2
89 6.64557 6 139 4.069767 4 189 2.932961 2
90 6.5625 6 140 4.038462 4 190 2.916667 2
91 6.481481 6 141 4.007634 4 191 2.900552 2
92 6.402439 6 142 3.977273 3 192 2.884615 2
93 6.325301 6 143 3.947368 3 193 2.868852 2
94 6.25 6 144 3.91791 3 194 2.853261 2
95 6.176471 6 145 3.888889 3 195 2.837838 2
96 6.104651 6 146 3.860294 3 196 2.822581 2
97 6.034483 6 147 3.832117 3 197 2.807487 2
98 5.965909 5 148 3.804348 3 198 2.792553 2
99 5.898876 5 149 3.776978 3 199 2.777778 2
100 5.833333 5 150 3.75 3 200 2.763158 2
101 5.769231 5 151 3.723404 3 201 2.748691 2
102 5.706522 5 152 3.697183 3 202 2.734375 2
103 5.645161 5 153 3.671329 3 203 2.720207 2
104 5.585106 5 154 3.645833 3 204 2.706186 2
105 5.526316 5 155 3.62069 3 205 2.692308 2
106 5.46875 5 156 3.59589 3 206 2.678571 2
107 5.412371 5 157 3.571429 3 207 2.664975 2
108 5.357143 5 158 3.547297 3 208 2.651515 2
109 5.30303 5 159 3.52349 3 209 2.638191 2
110 5.25 5 160 3.5 3 210 2.625 2
111 5.19802 5 161 3.476821 3 211 2.61194 2
112 5.147059 5 162 3.453947 3 212 2.59901 2
113 5.097087 5 163 3.431373 3 213 2.586207 2
114 5.048077 5 164 3.409091 3 214 2.573529 2
115 5 5 165 3.387097 3 215 2.560976 2
116 4.95283 4 166 3.365385 3 216 2.548544 2
117 4.906542 4 167 3.343949 3 217 2.536232 2
118 4.861111 4 168 3.322785 3 218 2.524038 2
119 4.816514 4 169 3.301887 3 219 2.511962 2
120 4.772727 4 170 3.28125 3 220 2.5 2
121 4.72973 4 171 3.26087 3 221 2.488152 2
122 4.6875 4 172 3.240741 3 222 2.476415 2
123 4.646018 4 173 3.220859 3 223 2.464789 2
124 4.605263 4 174 3.20122 3 224 2.453271 2
125 4.565217 4 175 3.181818 3 225 2.44186 2
126 4.525862 4 176 3.162651 3 226 2.430556 2
127 4.487179 4 177 3.143713 3 227 2.419355 2
128 4.449153 4 178 3.125 3 228 2.408257 2
129 4.411765 4 179 3.106509 3 229 2.39726 2
130 4.375 4 180 3.088235 3 230 2.386364 2
131 4.338843 4 181 3.070175 3 231 2.375566 2
132 4.303279 4 182 3.052326 3 232 2.364865 2
133 4.268293 4 183 3.034682 3 233 2.35426 2
134 4.233871 4 184 3.017241 3 234 2.34375 2
135 4.2 4 185 3 3 235 2.333333 2
136 4.166667 4 186 2.982955 2 236 2.323009 2

237 2.312775 2 287 1.895307 1 337 1.605505 1
238 2.302632 2 288 1.888489 1 338 1.60061 1
239 2.292576 2 289 1.88172 1 339 1.595745 1
240 2.282609 2 290 1.875 1 340 1.590909 1
241 2.272727 2 291 1.868327 1 341 1.586103 1
242 2.262931 2 292 1.861702 1 342 1.581325 1
243 2.253219 2 293 1.855124 1 343 1.576577 1
244 2.24359 2 294 1.848592 1 344 1.571856 1
245 2.234043 2 295 1.842105 1 345 1.567164 1
246 2.224576 2 296 1.835664 1 346 1.5625 1
247 2.21519 2 297 1.829268 1 347 1.557864 1
248 2.205882 2 298 1.822917 1 348 1.553254 1
249 2.196653 2 299 1.816609 1 349 1.548673 1
250 2.1875 2 300 1.810345 1 350 1.544118 1
251 2.178423 2 301 1.804124 1 351 1.539589 1
252 2.169421 2 302 1.797945 1 352 1.535088 1
253 2.160494 2 303 1.791809 1 353 1.530612 1
254 2.151639 2 304 1.785714 1 354 1.526163 1
255 2.142857 2 305 1.779661 1 355 1.521739 1
256 2.134146 2 306 1.773649 1 356 1.517341 1
257 2.125506 2 307 1.767677 1 357 1.512968 1
258 2.116935 2 308 1.761745 1 358 1.508621 1
259 2.108434 2 309 1.755853 1 359 1.504298 1
260 2.1 2 310 1.75 1 360 1.5 1
261 2.091633 2 311 1.744186 1 361 1.495726 1
262 2.083333 2 312 1.738411 1 362 1.491477 1
263 2.075099 2 313 1.732673 1 363 1.487252 1
264 2.066929 2 314 1.726974 1 364 1.483051 1
265 2.058824 2 315 1.721311 1 365 1.478873 1
266 2.050781 2 316 1.715686 1 366 1.474719 1
267 2.042802 2 317 1.710098 1 367 1.470588 1
268 2.034884 2 318 1.704545 1 368 1.46648 1
269 2.027027 2 319 1.699029 1 369 1.462396 1
270 2.019231 2 320 1.693548 1 370 1.458333 1
271 2.011494 2 321 1.688103 1 371 1.454294 1
272 2.003817 2 322 1.682692 1 372 1.450276 1
273 1.996198 1 323 1.677316 1 373 1.446281 1
274 1.988636 1 324 1.671975 1 374 1.442308 1
275 1.981132 1 325 1.666667 1 375 1.438356 1
276 1.973684 1 326 1.661392 1 376 1.434426 1
277 1.966292 1 327 1.656151 1 377 1.430518 1
278 1.958955 1 328 1.650943 1 378 1.42663 1
279 1.951673 1 329 1.645768 1 379 1.422764 1
280 1.944444 1 330 1.640625 1 380 1.418919 1
281 1.937269 1 331 1.635514 1 381 1.415094 1
282 1.930147 1 332 1.630435 1 382 1.41129 1
283 1.923077 1 333 1.625387 1 383 1.407507 1
284 1.916058 1 334 1.62037 1 384 1.403743 1
285 1.909091 1 335 1.615385 1 385 1.4 1
286 1.902174 1 336 1.610429 1 386 1.396277 1

387 1.392573 1 437 1.229508 1 487 1.100629 1
388 1.388889 1 438 1.226636 1 488 1.098326 1
389 1.385224 1 439 1.223776 1 489 1.096033 1
390 1.381579 1 440 1.22093 1 490 1.09375 1
391 1.377953 1 441 1.218097 1 491 1.091476 1
392 1.374346 1 442 1.215278 1 492 1.089212 1
393 1.370757 1 443 1.212471 1 493 1.086957 1
394 1.367188 1 444 1.209677 1 494 1.084711 1
395 1.363636 1 445 1.206897 1 495 1.082474 1
396 1.360104 1 446 1.204128 1 496 1.080247 1
397 1.356589 1 447 1.201373 1 497 1.078029 1
398 1.353093 1 448 1.19863 1 498 1.07582 1
399 1.349614 1 449 1.1959 1 499 1.07362 1
400 1.346154 1 450 1.193182 1 500 1.071429 1
401 1.342711 1 451 1.190476 1 501 1.069246 1
402 1.339286 1 452 1.187783 1 502 1.067073 1
403 1.335878 1 453 1.185102 1 503 1.064909 1
404 1.332487 1 454 1.182432 1 504 1.062753 1
405 1.329114 1 455 1.179775 1 505 1.060606 1
406 1.325758 1 456 1.17713 1 506 1.058468 1
407 1.322418 1 457 1.174497 1 507 1.056338 1
408 1.319095 1 458 1.171875 1 508 1.054217 1
409 1.315789 1 459 1.169265 1 509 1.052104 1
410 1.3125 1 460 1.166667 1 510 1.05 1
411 1.309227 1 461 1.16408 1 511 1.047904 1
412 1.30597 1 462 1.161504 1 512 1.045817 1
413 1.30273 1 463 1.15894 1 513 1.043738 1
414 1.299505 1 464 1.156388 1 514 1.041667 1
415 1.296296 1 465 1.153846 1 515 1.039604 1
416 1.293103 1 466 1.151316 1 516 1.037549 1
417 1.289926 1 467 1.148796 1 517 1.035503 1
418 1.286765 1 468 1.146288 1 518 1.033465 1
419 1.283619 1 469 1.143791 1 519 1.031434 1
420 1.280488 1 470 1.141304 1 520 1.029412 1
421 1.277372 1 471 1.138829 1 521 1.027397 1
422 1.274272 1 472 1.136364 1 522 1.025391 1
423 1.271186 1 473 1.133909 1 523 1.023392 1
424 1.268116 1 474 1.131466 1 524 1.021401 1
425 1.26506 1 475 1.129032 1 525 1.019417 1
426 1.262019 1 476 1.126609 1 526 1.017442 1
427 1.258993 1 477 1.124197 1 527 1.015474 1
428 1.255981 1 478 1.121795 1 528 1.013514 1
429 1.252983 1 479 1.119403 1 529 1.011561 1
430 1.25 1 480 1.117021 1 530 1.009615 1
431 1.247031 1 481 1.11465 1 531 1.007678 1
432 1.244076 1 482 1.112288 1 532 1.005747 1
433 1.241135 1 483 1.109937 1 533 1.003824 1
434 1.238208 1 484 1.107595 1 534 1.001908 1
435 1.235294 1 485 1.105263 1 535 1 1
436 1.232394 1 486 1.102941 1
We have all the information we need to perform our final calculations and
comparison. Let us summarize all the numbers using Table 20.
Table 20
Analysis of t(d)-values first digits
525 a
Digit Count Pk for t = d−10 ) Pk for y = x
(Count / Sum)
1 289 0.55576923 0.55555556
2 96 0.18461538 0.18518519
3 48 0.09230769 0.09259259
4 29 0.05576923 0.05555556
5 20 0.03846154 0.03703704
6 13 0.02500000 0.02645503
7 11 0.02115385 0.01984127
8 8 0.01538462 0.01543210
9 6 0.01153846 0.01234568
Sum 520 1.00000000 1.00000000
As we can see, the corresponding numbers are really close to each other. We
got such a nice result due to that fact that our application problem calculation
had 521 steps.
3. First digit probability of basic discrete functions
3.1. First digit probability of prime numbers
No formulas that were used by us to find the f dp values of a continuous function

can be used for a discrete function due to the way discrete functions are defined.
However, we still can use a technique that we applied for every example done
previously. Thus, we will build a table for x- and y-values and count the number
of each y-value’s first digit for every discrete function that is mentioned in this
chapter.
Let us recall that prime numbers are divisible only by the number 1 and
themselves.
A specific program, written in C + + programming language by the author
of this work, prints all prime numbers from 1 to 1000000 and then it prints the
report The number of primes from 1 to 1000000 is: 78498.
We will not print all of them in this work, however, the left side of Table 21
shows the beginning of calculations that are based on the report and the right
side of the same table shows the ending of the same calculations.
Table 21
Prime numbers and their first digits
Prime First Prime First Prime First Prime First
number digit number digit number digit number digit
2 2 233 2 998551 9 999221 9
3 3 239 2 998561 9 999233 9
5 5 241 2 998617 9 999239 9
7 7 251 2 998623 9 999269 9
11 1 257 2 998629 9 999287 9
13 1 263 2 998633 9 999307 9
17 1 269 2 998651 9 999329 9
19 1 271 2 998653 9 999331 9
23 2 277 2 998681 9 999359 9
29 2 281 2 998687 9 999371 9
31 3 283 2 998689 9 999377 9
37 3 293 2 998717 9 999389 9
41 4 307 3 998737 9 999431 9
43 4 311 3 998743 9 999433 9
47 4 313 3 998749 9 999437 9
53 5 317 3 998759 9 999451 9
59 5 331 3 998779 9 999491 9
61 6 337 3 998813 9 999499 9
67 6 347 3 998819 9 999521 9
71 7 349 3 998831 9 999529 9
73 7 353 3 998839 9 999541 9
79 7 359 3 998843 9 999553 9
83 8 367 3 998857 9 999563 9
89 8 373 3 998861 9 999599 9
97 9 379 3 998897 9 999611 9
101 1 383 3 998909 9 999613 9
103 1 389 3 998917 9 999623 9
107 1 397 3 998927 9 999631 9
109 1 401 4 998941 9 999653 9
113 1 409 4 998947 9 999667 9
127 1 419 4 998951 9 999671 9
131 1 421 4 998957 9 999683 9
137 1 431 4 998969 9 999721 9
139 1 433 4 998983 9 999727 9
149 1 439 4 998989 9 999749 9
151 1 443 4 999007 9 999763 9
157 1 449 4 999023 9 999769 9
163 1 457 4 999029 9 999773 9
167 1 461 4 999043 9 999809 9
173 1 463 4 999049 9 999853 9
179 1 467 4 999067 9 999863 9
181 1 479 4 999083 9 999883 9
191 1 487 4 999091 9 999907 9
193 1 491 4 999101 9 999917 9
197 1 499 4 999133 9 999931 9
199 1 503 5 999149 9 999953 9
211 2 509 5 999169 9 999959 9
223 2 521 5 999181 9 999961 9
227 2 523 5 999199 9 999979 9
229 2 541 5 999217 9 999983 9
Fig 41. Prime numbers in the range [1, 100)
Let us look at two graphs that show prime numbers. Figure 41 shows prime
numbers in the range [1, 100). The graph reminds us of the shape of an exponen-
tial function. It tells us that there is a possibility of smaller first digits having
higher f dp values.
Fig 42. Prime numbers in the range [1, 10000)
Figure 42 shows prime numbers in the range [1, 10000). This graph hints at
the same possibility; however, we should do our calculations to be able to state
this conclusion.
Let us calculate the probability of each first digit of prime numbers on the
ranges [1, 100), [1, 10000), and [1, 1000000). The results are shown in Table 22.
Table 22
Analysis of prime numbers f dp values
Digit f dp for [1, 100) f dp for [1, 10000) f dp for [1, 1000000)
1 0.16000000 0.13018714 0.12210502
2 0.12000000 0.11879577 0.11646157
3 0.12000000 0.11310008 0.11414304
4 0.12000000 0.11310008 0.11142959
5 0.12000000 0.10659072 0.10974802
6 0.08000000 0.10984540 0.10774797
7 0.16000000 0.10170871 0.10745497
8 0.08000000 0.10333605 0.10606640
9 0.04000000 0.10333605 0.10484344
Sum 1.00000000 1.00000000 1.00000000
The last column is more accurate than others due to its bigger range. It shows
that smaller first digits have higher probabilities than bigger ones, even though
the differences between them are not that big. The first two columns do not
show any stable rules due to that fact that their ranges are not big enough to
show any stability.
Let us look at the prime numbers f dp graph based on the range [1, 1000000).
The graph is shown in Figure 43. It is obvious that every Pk value is smaller
than the previous one. However, it does not look like the points of the graph
belong to a curve that could be described with an equation or at least with a
verbal description. We can just can make a hypothesis that if the range were
bigger, there would be a chance that we could find a way to describe the curve.
Fig 43. Prime numbers f dp based on the range [1, 1000000)

3.2. First digit probability of the Fibonacci function
Let us recall that the Fibonacci sequence is achieved by adding the last two
consecutive numbers. The first two terms of the sequence are both 1. In other
words, Fn = Fn−1 + Fn−2 when n ≥ 2, F0 = 1 and F1 = 1.
There is one more formula for defining each term of the sequence. Since we
will use it later, let us state it.
√
Φn −(−Φ)−n 1+ 5
Fn = √
5
, where Φ = 2 ≈ 1.618
A list of the first 500 Fibonacci numbers was created by using the first Fn
formula. The beginning and ending portions of the list are shown in Table 23.
Table 23
Fibonacci numbers and their first digits
n- Fn First n- Fn First n- Fn First
value value digit value value digit value value digit
1 1 1 35 9227465 9 467 1.77E+97 1
2 1 1 36 14930352 1 468 2.86E+97 2
3 2 2 37 24157817 2 469 4.63E+97 4
4 3 3 38 39088169 3 470 7.49E+97 7
5 5 5 39 63245986 6 471 1.21E+98 1
6 8 8 40 102334155 1 472 1.96E+98 1
7 13 1 41 165580141 1 473 3.17E+98 3
8 21 2 42 267914296 2 474 5.14E+98 5
9 34 3 43 433494437 4 475 8.31E+98 8
10 55 5 44 701408733 7 476 1.34E+99 1
11 89 8 45 1134903170 1 477 2.18E+99 2
12 144 1 46 1836311903 1 478 3.52E+99 3
13 233 2 47 2971215073 2 479 5.70E+99 5
14 377 3 48 4807526976 4 480 9.22E+99 9
15 610 6 49 7778742049 7 481 1.49E+100 1
16 987 9 50 12586269025 1 482 2.41E+100 2
17 1597 1 ... ... ... 483 3.90E+100 3
18 2584 2 ... ... ... 484 6.32E+100 6
19 4181 4 451 8.02E+93 8 485 1.02E+101 1
20 6765 6 452 1.30E+94 1 486 1.65E+101 1
21 10946 1 453 2.10E+94 2 487 2.68E+101 2
22 17711 1 454 3.40E+94 3 488 4.33E+101 4
23 28657 2 455 5.49E+94 5 489 7.01E+101 7
24 46368 4 456 8.89E+94 8 490 1.13E+102 1
25 75025 7 457 1.44E+95 1 491 1.83E+102 1
26 121393 1 458 2.33E+95 2 492 2.97E+102 2
27 196418 1 459 3.77E+95 3 493 4.80E+102 4
28 317811 3 460 6.09E+95 6 494 7.77E+102 7
29 514229 5 461 9.86E+95 9 495 1.26E+103 1
30 832040 8 462 1.60E+96 1 496 2.03E+103 2
31 1346269 1 463 2.58E+96 2 497 3.29E+103 3
32 2178309 2 464 4.18E+96 4 498 5.33E+103 5
33 3524578 3 465 6.76E+96 6 499 8.62E+103 8
34 5702887 5 466 1.09E+97 1 500 1.39E+104 1
Fig 44. Fibonacci numbers in the domain [1, 11]
Let us look at three graphs that show Fibonacci numbers. The graph in
Figure 44 shows the numbers in the domain [1, 11], the graph in Figure 45
shows the numbers in the domain [1, 16], and the graph in Figure 46 shows the
same function in the domain [1, 35].

As we can see, the Fibonacci function looks like a discrete approximation to

the exponential one. There is a reason for this and we will discuss it later. So
far, it looks like smaller digits have higher probabilities than bigger ones.
Let us prove it. We will calculate the probability of each first digit of the
Fibonacci function using our list of the first 500 Fibonacci numbers. Table 24
shows the Fibonacci f dp and the exponential f dp.
Table 24
Analysis of Fn values first digits
Digit Count f dp for y(n) = Fn f dp for y(x) = ax
1 151 0.30200000 0.30103000
2 88 0.17600000 0.17609126
3 63 0.12600000 0.12493874
4 47 0.09400000 0.09691001
5 40 0.08000000 0.07918125
6 33 0.06600000 0.06694679
7 29 0.05800000 0.05799195
8 27 0.05400000 0.05115252
9 22 0.04400000 0.04575749
Sum 500 1.00000000 1.00000000
The table above shows that smaller first digits have higher probabilities than
bigger ones.
Let us look at Figure 47, which shows the Pk values of the Fibonacci function
and the exponential one for a comparison. It is obvious that every Pk value of
the Fibonacci function is really close to the corresponding one of the exponential
function.
Fig 47. Pk values of the Fibonacci and exponential functions
3.3. First digit probability of the factorial function
Let us recall that x! = 1 · 2 · 3 · ... · (x − 1) · x, where x is a natural number. As it

was mentioned earlier, we are not using any formulas for discrete functions in
our calculations, only tables containing numbers.
A list of 2000 consecutive factorials was created [3]. Since the number 2000!
contains 5736 digits and the previous factorials are not much smaller, the expo-
nents of the calculated factorials were truncated every time they got too big.
We will print the beginning and the ending portions of the modified list in
Table 25.
Table 25
Factorials and their first digits
x- y = x! First x- y = x! First x- y = x! First
value digit value digit value digit
1 1 1 15 1.31E+12 1 29 8.84E+30 8
2 2 2 16 2.09E+13 2 30 2.65E+32 2
3 6 6 17 3.56E+14 3 31 8.22E+33 8
4 24 2 18 6.40E+15 6 32 2.63E+35 2
5 120 1 19 1.22E+17 1 33 8.68E+36 8
6 720 7 20 2.43E+18 2 34 2.95E+38 2
7 5040 5 21 5.11E+19 5 35 1.03E+40 1
8 40320 4 22 1.12E+21 1 36 3.72E+41 3
9 362880 3 23 2.59E+22 2 37 1.38E+43 1
10 3628800 3 24 6.20E+23 6 38 5.23E+44 5
11 39916800 3 25 1.55E+25 1 39 2.04E+46 2
12 479001600 4 26 4.03E+26 4 40 8.16E+47 8
13 6227020800 6 27 1.09E+28 1 ... ... ...
14 87178291200 8 28 3.05E+29 3 ... ... ...
x- y = x! First x- y = x! First x- y = x! First

value digit value digit value digit
1961 8.76E+206 8 1975 1.15E+253 1 1989 1.66E+299 1
1962 1.72E+210 1 1976 2.27E+256 2 1990 3.31E+302 3
1963 3.37E+213 3 1977 4.49E+259 4 1991 6.59E+105 6
1964 6.63E+216 6 1978 8.88E+262 8 1992 1.31E+109 1
1965 1.30E+220 1 1979 1.76E+266 1 1993 2.62E+112 2
1966 2.56E+223 2 1980 3.48E+269 3 1994 5.22E+115 5
1967 5.03E+226 5 1981 6.89E+272 6 1995 1.04E+119 1
1968 9.91E+229 9 1982 1.37E+276 1 1996 2.08E+122 2
1969 1.95E+233 1 1983 2.71E+279 2 1997 4.15E+125 4
1970 3.84E+236 3 1984 5.37E+282 5 1998 8.29E+128 8
1971 7.57E+239 7 1985 1.07E+286 1 1999 1.66E+132 1
1972 1.49E+243 1 1986 2.12E+289 2 2000 3.32E+135 3
1973 2.95E+246 2 1987 4.21E+292 4 ... ... ...
1974 5.82E+249 5 1988 8.37E+295 8 ... ... ...
Let us look at three graphs that show factorials. The graph in Figure 48
shows factorials in the domain [1, 5], the graph in Figure 49 shows factorials in
the domain [1, 10], and the graph in Figure 50 shows the same function in the
domain [1, 15].
Fig 48. Factorials in the domain [1, 5]
As we can see, the factorial function grows much faster than all the previous
ones. In addition, even though it would be hard to use our ruler for the f dp
comparison for this particular function, it looks like smaller digits have higher
probabilities than bigger ones and no ruler is needed for this conclusion.
Let us prove it now. We will calculate the probability of each first digit of
the factorial function using our list of the first 2000 factorials. Our results and
the exponential f dp values are shown in Table 26.
Table 26
Analysis of factorials first digits
Digit Count f dp for y(x) = x! f dp for y(x) = ax
1 591 0.29550 0.30103000
2 335 0.16750 0.17609126
3 250 0.12500 0.12493874
4 204 0.10200 0.09691001
5 161 0.08050 0.07918125
6 156 0.07800 0.06694679
7 107 0.05350 0.05799195
8 102 0.05100 0.05115252
9 94 0.04700 0.04575749
Sum 2000 1.00000 1.00000000
The table above shows that smaller first digits have higher probabilities than
bigger ones.
Let us look at Figure 51, which shows Pk values of the function y = x! and
the exponential function for a comparison.
Fig 51. Pk values of y = x! and y = ax
It is obvious that every Pk value of the factorial function is really close to the
corresponding one of the exponential function. It looks like that if our domain
were bigger, the Pk values would be even closer together. We will discuss this
hypothesis later.
4. Conclusion
We know by now that all functions that we have discussed have certain rules
for their Pk values. The values are either increasing, decreasing, or they stay
constant. Let us analyze how all of our functions are broken into these three
groups.
First, let us summarize everything that we have discovered in this work about
continuous functions. Our intuition tells us that behavior of Pk numbers has
something to do with the function’s shape. In particular, we should analyze
how the slope of a tangent to the function line affects its Pk numbers. We will
do this analysis for all our continuous functions by creating one picture per
group.
Figure 52 shows us graphs of the functions with decreasing Pk numbers. They
are y = 3x , y = x3 , and y = 50x . All of them represent their sections and one
representative per section should be enough.
50
Fig 52. Graphs of the functions y = 3x , y = x3 , and y = x
Since the function y = xa is decreasing on its entire domain if a is positive, it

is decreasing on the interval (0, a]. However, we are moving from a to 0 along
the x-axis while analyzing the y-values of this function. We are doing it because
our goal is to see how the first digits of y-values change while the y-values are
increasing. Thus, we have to move in the same direction while analyzing the
tangent lines of the function.
As we can see, tangent lines of all the functions in Figure 52 increase their
slopes while y-coordinates of their functions are increasing.
Let us look at another picture. The
√ functions with increasing Pk numbers are
shown in Figure 53. They are y = x and y = log2 x.
Again, one representative per section is enough to see the whole picture. It is
obvious that the slopes of the tangent lines are decreasing while the functions
y-coordinates are increasing.
√
Fig 53. Graphs of the functions y = x and y = log2 x
We should also cover the case when our function is linear. There is no need
to graph the function y = ax because it is obvious that all tangent lines will
have the same slopes as the function does due to the function’s nature.
Thus, we are ready to draw certain conclusions. Functions with decreasing Pk
numbers have tangent lines with increasing slopes and functions with increasing
Pk numbers have tangent lines with decreasing slopes. Moreover, functions with
constant Pk numbers have tangent lines with constant slopes. We know that
changes in directions of tangent lines are directly relevant to first and second
derivatives of the initial functions. Let us find them using our knowledge of
Calculus.
First, we will find the first and second derivatives of our continuous functions
and list the results below. Then, we will analyze whether the first derivatives
are increasing or decreasing and whether the second derivatives are positive or
negative. These analyses will be done on the chosen domain. The conclusions
are directly related to each other and to our Pk numbers.
f (x) = ax , a > 1, x ≥ 0
f 0 (x) = ax · ln a - increasing
f 00 (x) = ln a · ax · ln a = (ln a)2 · ax > 0
f (x) = xa , a > 1, x ≥ 1
f 0 (x) = axa−1 - increasing
f 00 (x) = a(a − 1)xa−1 > 0
1
f (x) = mx, m > 0, x ≥ m
f 0 (x) = m - constant
f 00 (x) = 0
√ 1
f (x) = a x = x a , a > 1, x ≥ 1 √
1 1 a
x
f 0 (x) = a1 · x a −1 = a1 · x a · x−1 = ax - decreasing
1−a 1−a
√
a
1 x·(1−a)
00 0 a −1
f (x) = ( a1 ·x a ) = 1
a · 1−a
a ·x = 1−a
a2 · x a −2 = a2 x2 <0
f (x) = loga x, a > 1, x ≥ a

f 0 (x) = ln 1a·x - decreasing
f 00 (x) = ( ln1a · x−1 )0 = (−1) · 1
ln a · x−2 = − ln a·x
1
2 < 0
f (x) = xa = ax−1 , a > 0, 0 < x ≤ a

f 0 (x) = (−1)ax−2 = − xa2 - decreasing
f 00 (x) = ((−1)ax−2 )0 = (−1)(−2)ax−3 = 2a
x3 >0
It is obvious that all functions with decreasing Pk numbers have positive

second derivatives and those with increasing Pk numbers have negative second
derivatives. In addition, the second derivative of the function with constant Pk
numbers equals to zero.
Let us graph all the Pk values of the functions with positive second derivatives
and the Pk numbers of a linear function on the same xy-plane for a comparison.
Figure 54 shows the graphs.
Fig 54. Pk values of a linear function and functions with positive second derivatives
As we can see, all the decreasing Pk functions change their values differently.
Let us analyze these differences.
We will use our knowledge of Calculus again to find the limits of the ratio
f (x+1)
f (x) for all the functions whose Pk numbers were just graphed. All the x-
variables of our limits will approach to positive infinity except the x-variable
of the limit of the reciprocal function. Since the x-values of the function are
decreasing while the y-values are increasing in our domain, the x-variable of
this function’s ratio’s limit will approach to zero. For the same reason, we will
f (x)
find the limit of the ratio f (x+1) for the reciprocal function instead of the one
mentioned above.
a
x x+1 1
lim = lim = lim (1 + ) = ∞;
x→0 a x→0 x x→0 x
x+1
ax+1
lim = lim a = a, a > 1;
x→∞ ax x→∞
(x + 1)a x+1 a 1
lim a
= lim ( ) = lim (1 + )a = 1a = 1;
x→∞ x x→∞ x x→∞ x
m(x + 1) x+1 1
lim = lim = lim (1 + ) = 1;
x→∞ mx x→∞ x x→∞ x
Let us summarize the results. It is ovbious that all exponential functions have
the same probabilities and all their ratios are equal to their bases, which are
greater than 1. Pk numbers of the reciprocal function start with bigger values
than those of the exponential functions and their ratios approach to infinity.
Pk numbers of the power functions start with smaller values than those of the
exponential ones and their ratios approach to 1. Thus, we can see a pattern that
higher limits belong to functions with higher P1 values.
Moreover, we might notice that Pk numbers of power functions with higher
exponents start with bigger values than those with lower exponents. However,
we do not know how high the Pk numbers will start if the exponent gets infinitely
big. Let us take another limit to answer this question. We will check what will
happen to the ratio f (x+1) a
f (x) , where f (x) = x if both x and a approach to infinity.
(x + 1)a x+1 a 1
lim = lim ( ) = lim (1 + )a = e;
a,x→∞ xa a,x→∞ x a,x→∞ x
The last step is valid due to the fact that [1]
1 x
lim (1 +
) = e;
x→∞ x
According to the result, we can make a hypothesis that as the exponent of a
power function gets bigger, its P1 approaches to P1 of an exponential function,
but never gets equal or bigger than it.
1 1 1
2 a −1 a 2 a −1
Based on the Pk formula of power functions, P1 = 1 = 1 , where a
10 a −1 10 a −1
1
2 x −1
is the function’s exponent. Let us graph the function y = 1 on the domain
10 x −1
[1, ∞) and see where it approaches to while x approaches to infinity. Figure 55

shows the graph.
1
2 x −1
Fig 55. The function y = 1 on the domain [1, ∞)
10 x −1
As we can see, the function mentioned above approaches to a number which

is close to the number 0.3 and never gets above it. Our intuition tells us that
this number is P1 of the exponential function, which is
1+1 ln 2
log10 = log10 2 = ≈ 0.301;
1 ln 10
1
Let us prove it algebraically. First, let x = c . Then, c → 0 as x → ∞.
1
2x − 1 2c − 1 (2c − 1)0 2c · ln 2 ln 2
lim 1 = lim = lim = lim = ;
x→∞ 10 − 1
x c→0 10c − 1 c→0 (10c − 1)0 c→0 10c · ln 10 ln 10
Thus, our assumption was correct. In addition, we can make a conclusion

that P1 values of the reciprocal, exponential, and power functions have their
boundaries and these boundaries do not overlap each other.
Let us look at the Pk graphs of functions with increasing Pk numbers. Those
functions have negative second derivatives. The Pk graph of a linear function
was added to the picture as well for a comparison. The graphs are shown in
Figure 56.
The Pk values of both logarithmic functions are shown for the range [1, 10). It
was proven before that these Pk numbers depend on the logarithmic function’s
range. It was also shown earlier that P9 values of the function are almost equal
to 1 and the rest of the Pk values are almost equal to 0 on all other ranges.
Thus, we are using the range mentioned above.
According to the graph, the P1 value of a root function with index a is greater
than the P1 value of a logarithmic function with base a. In addition, P1 of a
root function with higher index will be smaller than the one with lower index.
Moreover, P1 of a logarithmic function with higher base will be smaller than
the one with lower base.
Fig 56. Pk values of a linear function and functions with negative second derivatives
Let us look at limits of the ratios of the functions.
√ r r r
a
x+1 a x+1 a 1 a 1 √
lim √ = lim = lim 1 + = lim 1+ = a 1 + 0 = 1;
x→∞ a
x x→∞ x x→∞ x x→∞ x
ln(x+1)
loga (x + 1) ln a ln(x + 1) (ln(x + 1))0
lim = lim = lim = lim
x→∞ loga x x→∞ ln x
ln a
x→∞ ln x x→∞ (ln x)0
1 x
x+1 x x 1 1
= lim = lim = lim = lim = = 1;
x→∞ 1 x→∞ x + 1 x→∞ x
+ 1 x→∞ 1 + 1 1+0
x x x x
Thus, the ratios of both functions with increasing Pk numbers approach to 1.
We can just assume that the reason for it is the fact that P1 of those functions
is always smaller than P1 of the exponential one.
We had a chance to discuss how Pk numbers of the continuous functions
covered in the first chapter relate to each other. Summarizing all the mentioned
above relationships, we can conclude that if we have a real-life data set, there
is a possibility that we can find a continuous function that is a good fit for it
and has similar f dp numbers.
Let us talk about Pk numbers of the discrete functions discussed in the second
chapter. There are only three of them.
Pk values of prime numbers are slightly decreasing; however, the differences
between them are not big. The bottom part of its graph has a shape that reminds
us of an exponential or a power function. This fact might be an explanation of
why the Pk numbers are decreasing. However, if we graph the function on a
large domain, we will see that its shape has multiple small curves, but overall
it looks like a straight line. That is why the Pk values of the function are not
much different from each other.
Since there is no particular formula to define prime numbers, we will not do
any limits relevant to the function.
Let us talk about the Fibonacci function. Based on our calculations and
graphs, we made a hypothesis that the function follows Benford’s distribution.
Let us prove it by taking the limit of the ratio f (x+1)
f (x) , where f (x) = Fn .
√
Φn −(−Φ)−n 1+ 5
As it was mentioned above, Fn = √
5
, where Φ = 2 ≈ 1.618.
Φn+1 −(−Φ)−(n+1)
√
Fn+1 5 Φn+1 − (−Φ)−(n+1)
lim = lim Φn −(−Φ)−n
= lim
n→∞ Fn n→∞ √ n→∞ Φn − (−Φ)−n
5
Φn+1 − (−1)−(n+1) Φ−(n+1) Φn+1 − (−1)−(n+1) · Φn+1

1
= lim = lim
n→∞ Φn − (−1)−n Φ−n n→∞ Φn − (−1)−n · Φ1n
Φn (Φ − (−1)−(n+1) · Φ2n+1
1
) Φ − (−1)−(n+1) · Φ2n+1
1
= lim = lim
n→∞ Φn (1 − (−1)−n · Φ12 n ) n→∞ 1 − (−1)−n · Φ12 n
Φ−0
= = Φ;
1−0
Thus, the ratio of the Fibonacci function approaches to Φ and the function
can be approximated to a continuous exponential function y = Φx .
First, it explains why the Fibonacci function looks like an exponential one if
we graph it. In addition, it explains why their corresponding Pk numbers are
equal to each other.
Finally, we should talk about the function y = x!. Unfortunately, the limit of
the ratio f (x+1)
f (x) will not lead us to any meaningful results.
(x + 1)! x! · (x + 1)
lim = lim = lim (x + 1) = ∞;
x→∞ x! x→∞ x! x→∞
It may look like Pk numbers of the factorial function should be really different
from those of the exponential one; however, we know that this is not true. All of
our calculations show that Pk numbers of the factorial function approach and
get really close to Pk numbers of the exponential function on a large domain.
Thus, more work needs to be done to explain why the function y = x! has
Benford’s probability.
x √ √
According to Stirling’s formula [4], x! ≈ xex · 2πx = 2πx · ( xe )x .
Thus, the factorial function can be approximated to a continuous function,
which looks like an exponential
√ one with a particular coefficient. In other words,
x! ≈ max , where m = 2πx and a = xe . However, the coefficient and the base
of the function are not constant values, which makes the function different from
a regular exponential one.
Let us do a couple of limits. First, we will take a limit of the ratio m(x+1)
m(x)
a(x+1)
and then we will do a limit of a(x) where x approaches to infinity.
p √ r
m(x + 1) 2π(x + 1) x+1 x+1
lim = lim √ = lim √ = lim
x→∞ m(x) x→∞ 2πx x→∞ x x→∞ x
r
1 √
= lim 1 + = 1 + 0 = 1;
x→∞ x
x+1
a(x + 1) e x+1 1
lim = lim x = lim = lim (1 + ) = 1 + 0 = 1;
x→∞ a(x) x→∞
e
x→∞ x x→∞ x
Thus, even though the base and the coefficient depend on x, their ratios
approach to 1. It means that as x approaches to infinity, the base and the
coefficient become more stable.√
The function f (x) = x! ≈ 2πx · ( xe )x does not belong to a group of expo-
nential ones; however, it has the same Pk numbers as the exponential functions
do. There is a chance that the two ratios mentioned above shed some light on
this puzzle.
5. Acknowledgements
I would like to thank my son, Vitaliy Goncharenko, for his emotional support.
References
[1] Joyce D. (2012). e as the limit of (1 + 1/n)n , Clark University.

[2] Mcaskill B., Watt W., Balzarini E., Bonifacio L., Carlson S.,
Johnson B., Kennedy R., Wardrop H., Mackay S. (2011). Pre-
Calculus 11, 1st ed. McGraw-Hill
[3] Project Gutenberg EBook (2012). A List of Factorial Math Constants.
www.gutenberg.org
[4] Romik D. (2000). Stirling’s Approximation for n!: The Ultimate Short
Proof ? Tel-Aviv University, Israel
[5] Sisson P. (2008). College Algebra, 2nd ed. Hawkes Learning Systems
[6] Vital and Health Statistics (2002). 2000 CDC Growth Charts for the
United States: Methods and Development Series 11, No. 246, p 139, 145
[7] Walthoe J. (1999). Looking out for number one, www.plus.maths.org

First Digit Probability and Benford's Law

Uploaded by

Copyright:

Available Formats

First Digit Probability and Benford's Law

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

First Digit Probability and Benford's Law

Uploaded by

Copyright:

Available Formats

Probability Surveys

Vol. 0 (0000) 1–0

First Digit Probability and Benford’s

Received January 0000.

Fig 1. The measuring ruler for the range [1, 100).

Fig 2. The measuring ruler for the range [1, 10).

Fig 3. The measuring ruler for the range [10, 100).

Fig 4. The measuring ruler for the range [10n−1 , 10n ).

to derive a formula of the f dp of each function. As it was mentioned above, a

Let us use a general real-valued continuous function y = f (x) on the range

Fig 6. Continuous function y = f (x).

f −1 ((k + 1) · 10n−1 ) − f −1 (k · 10n−1 )

where k is a digit from 1 to 9 and n is a natural number. Since formula 2 is

2. Fdp of basic continuous functions

2.1. Pk of the exponential function y = ax

Let us first recall what the exponential function y = 2x looks like.

Fig 7. Exponential function y = 2x on the range [1, 10).

Fig 8. Exponential function y = 2x on the range [1, 1000).

Fig 9. Exponential function y = 3x on the range [1, 10).

Fig 10. Exponential function y = 3x on the range [1, 100000).

We had a chance to see the exponential functions with different bases on

Σ9k=1 Pk = Σ9k=1 (log10 (k + 1) − log10 k)

= (log10 (1 + 1) − log10 1) + (log10 (2 + 1) − log10 2) + (log10 (3 + 1) − log10 3)

+(log10 (4 + 1) − log10 4) + (log10 (5 + 1) − log10 5) + (log10 (6 + 1) − log10 6)

+(log10 (7 + 1) − log10 7) + (log10 (8 + 1) − log10 8) + (log10 (9 + 1) − log10 9)

= log10 2 − log10 1 + log10 3 − log10 2 + log10 4 − log10 3

+ log10 5 − log10 4 + log10 6 − log10 5 + log10 7 − log10 6

+ log10 8 − log10 7 + log10 9 − log10 8 + log10 10 − log10 9

Thus, Σ9k=1 Pk = 1, so we got one of the proofs.

Fig 12. Function y = log10 x.

N (t) - Final Amount of Bacteria

f −1 ((k + 1) · 10n−1 ) − f −1 (k · 10n−1 )

m loga h −m loga k·10h

m (loga h − loga k·10h )

Fig 13. Function N (t) = 300e0.04t .

Number of Amount of First Number of Amount of First

Number of Amount of First Number of Amount of First

Digit Count Frequency Benford’s

2.2. Pk of the power function y = xa

Fig 14. Quadratic function y = x2 on the range [1, 100).

Fig 15. Quadratic function y = x2 on the range [1, 1000).

Let us look at another graph of the function y = x2 . Figure 15 shows this

Fig 16. Cubic function y = x3 on the range [1, 10).

Fig 17. Cubic function y = x3 on the range [1, 1000).

Thus, Σ9k=1 Pk = 1, which satisfies one of the probabilitys law.

First, let y = x2 . Then a = 2 and

Fig 18. Pk values of the functions y = x2 and y = x3 .

Let us now do the same for the function y = x3 . √

In conclusion, we came back to the Formula 6. It means that the Pk values

Fig 21. The function D(t) = 16.1t2 .

Time in Distance First Time in Distance First Time in Distance First

Digit Count Frequency Pk values

2.3. Pk of the linear function y = mx

Fig 22. The function y = 5x.

The formula 2 will be used. Let f (x) = mx. Then, f −1 (x) = 1

Fig 23. Pk values of the function y = mx.

Time in Volume First Time in Volume First Time in Volume First

Time in Volume First Time in Volume First Time in Volume First

Digit Count Frequency Pk values