0% found this document useful (0 votes)
485 views13 pages

Wilcoxon Signed-Rank Test

The Wilcoxon Signed-Rank Test is a non-parametric alternative to the t-test for paired samples that does not assume a normal distribution. It was used to analyze the results of a study where 16 students provided probability ratings for two questions. Their differences in ratings were ranked and summed, yielding a test statistic W that was compared to critical values to determine if question A ratings tended to be higher than question B ratings, rejecting the null hypothesis of no difference.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
485 views13 pages

Wilcoxon Signed-Rank Test

The Wilcoxon Signed-Rank Test is a non-parametric alternative to the t-test for paired samples that does not assume a normal distribution. It was used to analyze the results of a study where 16 students provided probability ratings for two questions. Their differences in ratings were ranked and summed, yielding a test statistic W that was compared to critical values to determine if question A ratings tended to be higher than question B ratings, rejecting the null hypothesis of no difference.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 13

Wilcoxon Signed-Rank Test

A non-parametric alternative to the t-test


for independence samples.
Assumptions of the t-test
The scale of measurement for XA and XB
has the properties of an equal-interval
scale.
The differences between the paired values
of XA and XB have been randomly drawn
from the source population.
The source population from which these
differences have been drawn can be
reasonably supposed to have a normal
distribution.
Example
16 students in an introductory statistics course
are presented with a number of questions
concerning basic probabilities.
The question takes the form "What is the
probability of such-and-such?
The students are not allowed to perform
calculations. Their answers must be immediate,
based only on their raw intuitions. They are
instructed to frame each answer in terms of a
zero to 100 percent rating scale.
Continued
Reason: if students have developed a
good, solid understanding of the basic
concepts, they will tend to give higher
probability ratings for question A than for
question B.
Results
Subject XA XB XAXB
1 78 78 0
2 24 24 0
3 64 62 +2
4 45 48 3
5 64 68 4
6 52 56 4
7 30 25 +5
8 50 44 +6
9 64 56 +8
10 50 40 +10
11 78 68 +10
12 22 36 14
13 84 68 +16
14 40 20 +20
15 90 58 +32
16 72 32 +40
meandifference=+7.75
Continued
The observed results are consistent with
the hypothesis.
The probability ratings do on average end
up higher for question A than for
question B.
Determine whether the degree of the
observed difference reflects anything more
than due to chance.
The Wilcoxon Test
1 2 3 4 5 6 7
rankof
absolut
original absolute e signed
Subject XA XB XAXB XAXB XAXB rank
1 78 78 0 0 --- ---
2 24 24 0 0 --- ---
3 64 62 +2 2 1 +1
4 45 48 3 3 2 2
5 64 68 4 4 3.5 3.5
6 52 56 4 4 3.5 3.5
7 30 25 +5 5 5 +5
8 50 44 +6 6 6 +6
9 64 56 +8 8 7 +7
10 50 40 +10 10 8.5 +8.5
11 78 68 +10 10 8.5 +8.5
12 22 36 14 14 10 10
13 84 68 +16 16 11 +11
14 40 20 +20 20 12 +12
15 90 58 +32 32 13 +13
16 72 32 +40 40 14 +14
W=67.0
T
N=14
Logic and procedure
sum=N(N+1)/2=14(14+1)/2= 105
The maximum possible positive value
of W (in the case where all signs are
positive) is W=+105
The maximum possible negative value (in
the case where all signs are negative) is
W=105.
Continued (1)
The null hypothesis is that there is no
tendency in either direction, hence that the
numbers of positive and negative signs
will be approximately equal.
In that event, we would expect the value
of W to approximate zero, within the limits
of random variability.
Hence W = 0.
Continued (2)
For any particular value of N, it can be
shown that the standard deviation of the
sampling distribution of W is equal to

W = sqrt[N(N+1)(2N+1)/6]

For the present example, with N=14,

W = sqrt[14(14+1)(28+1)/6]= 31.86
Continued (3)
the structure of the z-ratio for the Wilcoxon
test is
(WW) 0.5
z=
W 67 - 0.5
For the example: z = = +2.09
31.86
Continued (4)
= 0.05.
Critical value of z for two-tailed test =
1.96.
Reject the null hypothesis.
Critical values for small samples

One tailed test
0.05 0.025 0.01 0.005
Two tailed test
N -- 0.05 0.02 0.01
5 15 -- -- --
6 17 21 -- --
7 22 24 28
8 26 30 34 36
9 29 35 39 43

You might also like