STA210 | HYPOTHESIS TESTING STA210 | HYPOTHESIS TESTING

(NON-PARAMETRIC METHODS) (NON-PARAMETRIC METHODS)

5.1 Introduction to Non-parametric Methods 5.2 Sign Test (Binomial Test when )
• Used to test hypotheses on a population median
Ø Also known as “distribution free methods”. • Assumptions:
Ø Used when the normality assumption cannot be fulfilled and also other ü Random sample
assumptions that validate the use of parametric methods. ü Measurement scale at least ordinal
Ø Example:
ü Two judges rank five brands of detergents by assigning a rank of 1 to Sign Test for One sample / Two related sample (dependent)
the brand believed to have the best overall quality, a rank of 2 to the
second best, and so forth. Finally we want to see whether is there any
Left tailed Right tailed Two tailed
agreement between the two judges. Step 1: H0
1
Detergent 🧴 🧴 🧴 🧴 🧴 sample
Step 1:
H1
H0
Ranking by 2
Judge A samples H 1
Step 2: TS
Ranking by Binomial variable X with (written as )
Judge B
*p is probability of having positive sign
x = number of “+” x = number of “+” x = number of “+” sign or “-”
5.1.1 Conditions for using Non-parametric Methods sign sign sign, whichever is smaller
p-value

Ø Small sample size, unless the actual distribution is known exactly.

Ø The data are measure in categorical manner (nominal/ordinal scale).
Ø For the data in ordinal or interval scale, the distribution function is unspecified
or other than normal. Step 3: CV
Step 4: DR Reject H 0 if p-value
Step 5: Make a decision and conclusion
5.1.2 Advantages and Disadvantages of Non-parametric Methods

Advantages Disadvantages
May be used on all types of Less efficient than parametric methods when both
data (categorical/ratio) method are applicable.
Easy to apply for small sample When the sample size are large, the calculation
sizes become more tedious
Less assumption needed Not utilize all the information in the sample

• If a parametric and a nonparametric test are both applicable to the same set of
data, we should carry out the more efficient parametric technique.
• On certain situation, when some of the assumptions cannot be met such
normality, then the non-parametric procedure much more efficient than the
parametric.

139 141
 STA210 | HYPOTHESIS TESTING STA210 | HYPOTHESIS TESTING
(NON-PARAMETRIC METHODS) (NON-PARAMETRIC METHODS)

5.3 Sign Rank Test (also known as Wilcoxon Signed Rank Test) - Two related
sample (dependent)
Example 2
The following samples of six data, the lengths of a particular plant in mm were recorded
• Assumptions
from an experiment:
ü Random sample/s
1.0, 1.5, 1.8, 1.9, 1.4, 2.0 ü Measurement scale at least interval
At 5% significance level, use a sign test to investigate whether the median length of Procedures:
the population of plants is greater than 1.5mm
For one sample:
Solution
1. Subtracts from each sample value, discarding all the differences that equals
1.0 1.5 1.8 1.9 1.4 2.0 to zero
2. Rank the differences by ignoring the sign (from smallest to largest)
3. If the absolute value of two or more differences is the same, assign to each the
average of the ranks
For two dependent sample, testing

1. Calculate the differences for each pair, subtracts from each differences
2. Rank the differences by ignoring the sign (from smallest to largest)

Calculate
o : Sum of ranks to the positive difference
o : Sum of ranks to the negative difference
o : The smaller of or

Refer Table 17 to find the CV

Left tailed Right tailed Two tailed
Step 1: H0
1
sample H1
Step 1: H0
2
samples H 1
Step 2: TS

Step 3: CV

Step 4: DR Reject H 0 if Reject H 0 if Reject H 0 if

Step 5: Make a decision and conclusion

143 145
 STA210 | HYPOTHESIS TESTING STA210 | HYPOTHESIS TESTING
(NON-PARAMETRIC METHODS) (NON-PARAMETRIC METHODS)

Example 5 Example 6
A paired experiment was conducted to compare two populations. The data is shown The nicotine content of two brands of cigarettes, measured in milligrams, was found
in the following table. Use a signed-rank test to verify the difference between two to be as follows:
population. Use = 0.05.
Sorting the data sets from smallest to largest
Pairs Brand A Brand B Data Rank
population 2.1 4.1 0.6
1 2 3 4 5 6 7
4.0 0.6 1.6
1 8.9 8.1 9.3 7.7 10.4 8.3 7.4 6.3 3.1
1.9
2 8.8 7.4 9 7.8 9.9 8.1 6.9 5.4 2.5
2.1
4.8 4.0
3.7 6.2 2.2
Rank
6.1 1.6 2.5
3.3 2.2 3.1
Solution 1.9 3.3
5.4 3.7
4
4
4.1
4.8
5.4
5.4
6.1
6.2
6.3

Test the hypothesis, at the 0.05 level of significance, that the median nicotine contents
of the two brands are equal against the alternative that they are unequal.
Solution

147 149
 TUTORIAL 5 | ANALYSIS OF CATEGORICAL DATA TUTORIAL 5 | ANALYSIS OF CATEGORICAL DATA

UPDATED NON-PARA TEST At 5% level of significance, do the data provide sufficient evidence to conclude
that the average number of rainy days is less than 124.9 days? Use signed
1. The following data represent the time, in minutes, that a patient has to wait rank test.[w+=3,reject]
during 12 visits to a doctor’s office before being seen by the doctor:
6. [Mar 14] A sample of 9 athletes, chosen at random , recorded the following
number of push ups that they completed before and after a stamina-building
17 15 20 20 32 28
programme.
12 26 25 25 35 24 Athlete Before After
1 25 30
Use the sign test at the 0.05 level of significance to test the doctor’s claim that 2 14 16
the median waiting time for her patients is not more than 20 minutes. 3 20 20
[x=7,p-value=0.1719] 4 17 16
5 28 27
2. The aggressiveness scores of 12 sets of identical twins are given below. Use 6 9 11
the sign test at the 0.05 level of significance to test the null hypothesis that the 7 36 38
Twin I is not less aggressive than Twin II. 8 26 29
9 30 33
Twin set Aggressiveness score twin I Aggressiveness score twin II Use the Wilcoxon method to test at the 0.05 level of significance whether the
number of push ups has increased after a stamina-building programme.
1 82 84 [w+=3,reject]

2 78 78 7. A bulb manufacturer claims that brand A bulbs last longer than brand B bulbs.
To test this claim, the life of both brands of bulbs, per hour, were recorded:
3 82 84
Brand A 810 850 760 690 820 830
4 77 79
Brand B 765 720 700

5 83 80
Use the rank-sum test with α = 0.05 to test whether the claim is valid. [u1=4,
78 84 claim not valid]
6
86 88 8. From a science class of 12 equally capable students preparing for an entrance
7
examination, 5 students are selected at random and given additional coaching
8 82 84 by an expert. Their grades are noted:

9 70 73 Grades of students
who receive 85 75 72 81 90
10 69 78 additional coaching
Grades of students
11 79 78 who do not receive 67 74 84 79 93 88 68
additional coaching
12 65 65
Use the rank-sum test with α = 0.05 to determine if the additional coaching
affects the average grade of students in the entrance examination. [u=15, fail
to reject]

147 149