Coin Probability
Coin Probability
Coin Probability
Dashboard
Login
Dashboard
Calendar
Inbox
History
Help
Support Resources
Data
Pages
P7: Coin-Tossing Distributions
Home
Modules
WebEx
The below figure shows the path of four balls that fall through a Galton board. The
chances of falling in the locations follow a special probability distribution that has a
strong connection with a simple coin-tossing experiment.
Preview
Consider the following random experiment. You take a quarter and flip it ten times,
recording the number of heads you get. There are four special characteristics of this
simple coin-tossing experiment.
1. You are doing the same thing (flip the coin) ten times. We will call an individual
coin flip a trial, and so our experiment consists of ten identical trials.
2. On each trial, there are two possible outcomes, heads or tails.
3. In addition, the probability of flipping heads on any trial is ½.
4. The results of different trials are independent. This means that the probability of
heads, say, on the fourth flip, does not depend on what happened on the first three
flips.
We are interested in the number of heads we get – we will refer to this number as X.
In particular, we are interested in the probability of getting five heads, or Prob(X = 5).
In this topic, we will see that this binomial probability model applies to many different
random phenomena in the real world. We discuss probability computations for the
binomial and closely related negative binomial models and illustrate the usefulness
of these models in representing the variation in real-life experiments.
In this topic, the learning objectives are to:
· Understand how real coin tossing can be differentiated from fake coin tossing
where students are imagining flips of heads and tails.
NCTM Standards
Trial 1 2 3 4 5 6 7 8 9 10
Resul
HHTTHTTHHT
t
Another possible outcome is TTHHTHTHHH. The sample space consists of all
possible ordered listings of ten letters, where each letter is either an H or a T.
Next, consider computing the probability of a single outcome of ten flips such as the
HHTTHHTHHT sequence shown above. We can write the probability of this outcome
as
Prob(“H on toss 1” AND “H on toss 2” AND “T on toss 3” AND … AND “T on toss
10”)
Using the fact that outcomes on different trials are independent, this probability can
be written as the product
Prob(H on toss 1) x Prob(H on toss 2) x Prob(T on toss 3) x … x Prob(T on toss 10).
Since the probability of heads (or tails ) on a given trial is ½, we have
Trial 1 2 3 4 5 6 7 8 9 10
Result
If we observe five heads, then we wish to place five H’s in the ten slots above. In the
outcome HHHHHTTTTT, the heads occur in trials 1, 2, 3, 4, 5, and in the outcome
HHHTTTTTHH, the heads occur in trials 1, 2, 3, 9, and 10. If we observe exactly 5
heads, then we must choose five numbers from the possible trial numbers 1, 2, …,
10 to place the five H’s. There are 10 C 5 ways of choosing these trial numbers. (The
order in which we choose the trial numbers is not important.) Since there are 10 C 5
ways of getting exactly five heads, and each outcome has probability 1/2^(10) , we
see that
Prob(X = 5) = (10 C 5) 1/2^10 = 0.246.
From a basic property of probabilities, we see that the Prob(five heads are
not tossed) = 1- 0.246 = 0.754. It is interesting to note that although we expect to get
five heads when we flip a coin ten times, it is actually much more likely not to flip five
heads than to flip five heads.
X 0 1 2 3 4 5
P(X)
4. By using counting arguments, how many sequences of five flips will contain
exactly 2 heads? Check that this number agrees with your computation in part 3.
5. If you flip a coin 20 times, how many outcomes will result in exactly 10 heads?
What is the probability of 10 heads?
Binomial Experiments
Although the coin tossing experiment described above seems pretty artificial, many
random experiments share the same basic properties as coin tossing. Consider the
following binomial experiment:
1. We are repeating the same basic task or trial many times – let the number of trials
be denoted by n.
2. On each trial, there are two possible outcomes, which we will call “success” or
“failure”. (We could call the two outcomes “black” and “white”, or “0” or “1”, but they
are usually called success and failure.)
3. The probability of a success, denoted by p, is the same for each trial.
4. The results of outcomes from different trials are independent.
Here are some examples of binomial experiments.
A sample survey . Suppose the Gallup organization is interested in estimating the
proportion of adults in the United States who use the popular auction web site EBay.
They take a random sample of 100 adults and 45 say that they use EBay. In this
story, we see that
At this point, things are looking good – this may be a binomial experiment. But …
Pretend Coin Flips X Y Z
2. (Real coin tossing) Now flip a quarter 200 times – put your results (H or T) in the
boxes below.
Real Coin Flips X Y Z
5. Compare the number of heads for the fake coin flips and the real coin flips by
constructing parallel dotplots. By looking at the two histograms and calculating
suitable summary statistics, explain how the numbers of heads for the real coins look
different from the number of heads for the fake coins.
6. Do the same comparison using the longest run variable. Which dataset tends to
have “long” runs – the fake coins or the real coins?
7. Repeat the comparison using the number of switches variable. Do you notice any
differences between the histogram for the number of switches for the real coins and
the number of switches for the fake coins?
2. Select the Collection and simulate 20 flips by typing Apple (or Control) – Y. For
each simulation, record the number of heads, and the length of the longest run.
1
2
3
4
5
3. This Fathom program has been set up to repeat this experiment (of flipping 20
coins ) 1000 times – for each experiment, we record the number of heads and the
length of the longest run.
To do this, select the Measures from Collection and type Apple-Y. This collection
contains the number of heads and length of longest run for these 1000 experiments.
4. First look at the Number of Heads for the 1000 experiments.
(a) Construct a histogram of the number of heads.
(b) Construct a count or frequency table of the number of heads.
(c) What is the most likely number of heads you will get?
(d) What is the probability you will flip exactly 10 heads?
(e) What is the probability you will flip 15 or more heads?
5. Now look at the Length of the Longest Run for the 1000 experiments.
(a) Construct a histogram and count table.
(b) What is the most likely length of the longest run?
(c) What is the probability the longest run is 6 or more?
6. Above, we assumed that p = .5 (we were flipping a fair coin). Now using the slider,
change p to .3. Redo the simulation of 1000 experiments.
Answer the same questions as in 4 and 5.
(8:06) Bryant Jump Shot: MADE (11:15) Bryant Turnaround Jump: MADE
(7:33) Bryant Jump Shot: MADE (10:31) Bryant Jump Shot: MISSED
(6:46) Bryant Jump Shot: MADE (8:31) Bryant Layup Shot: MISSED
(5:50) Bryant Jump Shot: MADE (7:04) Bryant Jump Shot: MISSED
(5:16) Bryant Jump Shot: MISSED (6:33) Bryant Jump Shot: MISSED
(4:55) Bryant Dunk Shot: MADE (4:23) Bryant Driving Finger Roll: MADE
(4:21) Bryant Turnaround Jump: MISSED (2:06) Bryant Dunk Shot: MADE
(3:15) Bryant Jump Shot: MISSED (0:57) Bryant Jump Shot: MISSED
(0:39) Bryant Jump Shot: MISSED (8:19) Bryant Turnaround Jump: MADE
(0:01) Bryant Jump Shot: MISSED (6:54) Bryant Slam Dunk Shot: MADE
(9:41) Bryant Jump Shot: MISSED (4:12) Bryant Jump Shot: MISSED
(5:20) Bryant Jump Shot: MADE (1:57) Bryant Jump Shot: MADE
(4:38) Bryant Jump Shot: MADE (0:47) Bryant Fade Away: MISSED
(4:03) Bryant Fade Away: MISSED (0:18) Bryant Jump Shot: MISSED
(0:02) Bryant Turnaround Jump: MISSED (4:26) Bryant Turnaround Jump: MISSED
1. For Kobe’s data, compute the length of the longest run of makes or misses.
2. Do you think this value is unusually small or large? Why?
3. One way of deciding if Kobe’s longest run of makes or misses is unusual is to
compare this value with the longest run of heads or tails in 40 flips of a fair coin. (We
use a fair coin since Kobe’s probability of making a particular shot is approximately .
5.) Below we have simulated 20 sequences of forty coin flips. For each sequence,
compute Y = the longest run of heads or tails and record this value on the right.
Y = longest run
seq 1 HTHTHHHTTTTTHHTHTHTTHHHTTTHTHTTHHTTHTTTH
seq 2 TTHTHHTTTTTTTHTHHTHTTHTHHTHTTHTTTTHHHTHT
seq 3 TTHHTHHTHTHHTHTHTTTHTTTHHHHTHHHHHTHTHTTH
seq 4 THHTTHHHHTTTTHHTTTTHHTTTTHHTHTHHHHHTHTHH
seq 5 HHTTTTHHHHHHTHTHHHTHTHHHTTHHTTHHHHTTHHTT
seq 6 THTHHHTTHHTTTTTTHHHHHHTHHTHTHTTHTHTHHHHH
seq 7 HHHHHTHHHHTTTHHTHTHTHHTHTTHHHHTTHTHTHHTT
seq 8 HTHHTTTHHTTTHTTTHTHHHHTTHTHTHTTHHTTHTTHH
seq 9 TTHHHHHHHTTHHTTTTTHHTHHTHHHTTHHTHTTHHHHT
seq 10 HTHTTTTHTHTTHTHTHTHTTHHHTHTTTHHHHHHTTTTT
seq 11 HTTTTTHTHHTTTHTHHHHHHHHHTHHTTTHTHTTTHHTH
seq 12 TTHHHTHTHHHTHHHTHTHHHTHHTTTHHHHHHTTTTTHT
seq 13 THHTTHTHTHHTHHHTHHHTTHTTHTTTHTTHTTHHHTHH
seq 14 TTTTHHHHTTHTTTHTHTHTHTTTHTTHTTHTTHTTHTTH
seq 15 THHHTHTTTTHTTTTTTHHHHHHTHTTHTHHHHHTTTHTH
seq 16 THHHTTTHHTHHHTHTHTTHHHHHHHTTHTHTHHHHHTTH
seq 17 HHTHHTTTTHHHHTTHTTTTHHHHHTTTHHTHHTHHHHHT
seq 18 THHTTTTTTHHHTTHTTHHHTTTHHHTHTTTHTTHTHHTH
seq 19 THHTTTHTTHTTHHTHHTTTTHHHHTTHTTTTHHHTTHHT
seq 20 HTTTHHTHTTTTTHTHHHTHTHTTTHTHHTTTHTTHTTTH
4. Construct a dotplot of the 20 values of the longest run Y. Indicate the length of
Kobe’s longest run by a vertical line placed on the dotplot. Based on this graph,
would you say that Kobe’s longest run is unusual relative to the distribution of the
longest run for 40 coin flips? Explain.
Binomial Computations
A binomial experiment is defined by two numbers
n = the number of trials, and p = probability of a “success” on a single trial.
If we recognize an experiment as being binomial, then all we need to know is n and p
to determine probabilities for the number of successes X.
Using the same argument as we made in the coin-tossing example, one can show
that the probability of k successes in a binomial experiment is given by
P(k) = (n C k) p^k (1 - p)^(n - k)
Let’s illustrate using this formula for a few examples.
Baseball example (revisited). Remember our baseball player with a true batting
average of .300 is coming to bat five times during a game. What is the probability
that he gets exactly two hits?
We showed earlier that this was a binomial experiment. Since the player has five
opportunities, the number of trials is n = 5. If we regard a success as getting a hit,
the probability of success on a single trial is p = 0.3. The random variable X is the
number of hits of the player during this game.
Using the formula, the probability of exactly two hits is
P(2) = (5 C 2) (.3)^2 (1 - .3)^3 = .309.
What is the probability that the player gets at least one hit? To do this problem, we
first construct the collection of binomial probabilities for n = 5 trials and probability of
success p = 0.3. The table below shows all possible values of X (0, 1, 2, 3, 4, 5) and
the associated probability that can be found using the binomial formula.
X P(X)
0 0.168
1 0.360
2 0.309
3 0.132
4 0.029
5 0.002
We are interested in the probability that the player gets at least one hit or Prob(X >=
1). “At least one hit” means that X can be 1, 2, 3, 4, or 5. To find this we simply sum
the probabilities of X between 1 and 5:
Prob(X>=1) = P(X = 1, 2, 3, 4, 5) = 0.360 + 0.309 + 0.132 + 0.029 + 0.002 = 0.832.
There is a simpler way of doing this computation using the complement property of
probabilities. We note that if the player doesn’t get at least one hit, then he was
hitless in the game (that is, X = 0). Using the complement property
Prob(X>=1) = 1 – Prob(X = 0) = 1 – 0.168 = 0.832.
X P(X)
0 0.016
1 0.094
2 0.234
3 0.312
4 0.234
5 0.094
6 0.016
Y P(y)
2 .1384
3 .1738
4 .1637
5 .1371
6 .1076
7 .0811
8 .0594
9 .0426
Note that it is most likely that Ichiro will only need three at-bats to get his two
additional hits, but the probability of three at-bats is only 17%. Actually each of the
values 2, 3, 4, 5, and 6 have probabilities exceeding 10%. There is a significant
probability that Ichiro will take a large number of bats – by adding the probabilities in
the table, we see that the probability that Y is at most 9 is .904, so the probability that
Y exceeds 9 is 1 - .904 = .096.
For a negative binomial experiment where Y is the number of trials needed to
observe r successes, one can show that the mean value is
E(Y) = r / p .
For our baseball example, r = 2 and p = .372, so the expected number of at-bats to
get two hits would be E(Y) = 2/.372 = 5.4. It is interesting to note that although Y = 3
is the most probable value, Ichiro would average over 5 at-bats to get 2 hits in many
repetitions of this random experiment.
1. Record the number of makes (successes) for each game – put your answers in
the blanks.
2. Collect the number of makes from all students in. Construct a suitable graph of
these data.
3. What is the most likely number of successes during a game of 10 shots?
4. Find the probability she makes at least half her shots.
5. Find the probability she makes all of her shots.
6. Suppose that the shooter continues to shoot free throws until she misses three
shots. Use this same diagram to record the results of the individual shots until the
experiment is completed. For each experiment, record Y = the total shots taken.
Game 1: Total shots = ______ Game 2: Total shots = ______
Game 1: Total shots = ______ Game 2: Total shots = ______
7. Collect the number of makes from all students in. Construct a suitable graph of
these data.
8. What is the most likely number of shots taken?
9. What is the probability the woman will take at least 8 shots?
Wrap-up
In this topic, we were introduced to the binomial experiment which represents a
popular type of random experiment that resembles coin tossing. The experiment is a
sequence of trials where there are two possible outcomes on each trial, the
probability of a success is the same for each trial, and outcomes from different trials
are independent. The focus is on the number of successes X that has a binomial
distribution with parameters n, the number of trials, and p, the probability of success.
A probability formula to compute P(X = k) was derived, and simple expressions were
presented for the mean and standard deviation of X. The negative binomial
experiment is a similar coin-tossing experiment where one continues sampling until
one observes r successes and the random variable is Y, the number of trials.
EXERCISES
1. Binomial Experiments
Is each random process described below a binomial experiment? If it is, give values
of n and p. Otherwise, explain why it is not binomial.
a. Roll a die 20 times and count the number of sixes you roll.
b. There is a room of 10 women and 10 men – you choose five people from the room
without replacement and count the number of women you choose.
c. Same process as (b) but you sample with replacement instead of without
replacement.
d. You flip a coin repeatedly until you observe 3 heads.
e. The spinner shown to the right is spun 50 times – you count the number
X 0 1 2 3 4 5 TOTAL
Count 10 29 44 40 15 3 141
To help understand this table, note that the count for X=1 is 29 – this means there
were 29 periods where Abreu was on-base exactly one time. The count for X=2 is 44
– this means that for 44 periods Abreu was on-base two times.
Since each outcome is either a success or failure, where success is getting on-base,
one wonders if the variation in these data can be explained by a binomial
distribution.
X 0 1 2 3 4 5 TOTAL
P(X)
Expected
Count
a. Find the probabilities for a binomial distribution with n=5 and p=.443. (This value
of p is Abreu’s on-base rate for the entire 2004 baseball season.) Place these
probabilities in the P(X) row of the table.
b. Multiply the probabilities you found in part (a) by 141, the number of periods in the
2004 season. Place these numbers in the Expected Count row of the table. These
represent the expected number of times Abreu would have 0, 1, 2, .., 5 times on-
base if the probabilities followed a binomial distribution.
c. Compare the expected counts with the actual observed counts in the first table.
Does a binomial distribution provide a good description of these data?
16. Graphs of Binomial Distributions
The below figures show the binomial distributions with n = 20 and p = .5 (left) and n
= 20 and p = .2 (right).
Recall the 68 rule from Topic P6 that said that if a probability distribution is
approximately bell-shaped, then approximately 68% of the probability falls within one
standard deviation of the mean.
a. For the binomial distribution with n = 20 and p = .5, find the mean μ and
standard deviation σ and compute the interval ( μ – σ, μ + σ).
b. Find the exact probability that X falls in the interval ( μ – σ, μ + σ).
c. Repeat parts a and b for the binomial distribution n = 20 and p = .2.
d. For which distribution was the 68% rule more accurate? Does that make sense
based on the shapes of the two distributions?
17. Guessing on a Test
Students in a statistics class were given a five-question baseball trivia quiz. On each
question, the students had to choose one of two possible answers. The number
correct X was recorded for each student – a count table of the values of X is shown
below.
0 0
1 3
2 4
3 7
4 6
5 1
a. Suppose the students know little about baseball and so they are guessing on each
question. If this is true, find the probability distribution of the number correct X.
b. Using this distribution, find the probability of each value of X and place these
probabilities in the above table.
c. By multiplying these probabilities by the number of students (21), find the
expected number of students for each value of X.
d. Compare your expected counts with the actual counts – does a binomial
distribution seem like a reasonable assumption in this example?
18. Playing Roulette
Suppose you play the game roulette 20 times. Each game, you place a Trio Bet on
three numbers and you win with probability 3/38.
a. Find the probability you win the game exactly two times.
b. Find the probability that you are winless in the 20 games.
c. Find the probability you win at least once.
d. How many games do you expect to win?
19. The Galton Board
Consider the Galton board described in the Spotlight at the beginning of this topic. A
ball is placed above the first peg and dropped. When it strikes a peg, it is equally
likely to fall left or right. The location at the bottom X is equal to the number of times
that the ball falls right.
a. Explain why X has a binomial distribution and give the values of n and p.
b. Find P(X = 2).
c. Find the probability the ball falls to the right of the location “1”.
d. Suppose that we change the experiment so that the probability of falling right is
equal to 1/4. Explain how this changes the binomial experiment and find P(X = 2).
20. Drug Testing
In a New York Times article “Facing Questions, Rodriguez Raises More” (February
21, 2008), Major League Baseball is said to have a drug-testing policy where 600
tests are randomly given to a group of 1200 professional ballplayers. Alex Rodriguez
claimed one season that he received five random tests.
a. If every player is equally likely to receive a single random blood test, what is the
probability that Rodriguez gets tested?
b. If X represents the number of tests administered to Rodriguez among the 600 test,
then explain why X has a binomial distribution and give the values of n and p.
c. Compute the probability that Rodriguez receives exactly one test.
d. Recall Rodriguez’s claim that he received five random tests. Compute the
probability of this event.
e. You should find the probability computed in part d to be very small. If Rodriguez is
indeed telling the truth, what do you think about the randomness of the drug-testing
policy?