Dr. Md. Abdus Salam Akanda Website: http://du.ac.

bd
Professor of Statistics, DU E-mail: [email protected]

Non-parametric Tests

Occasionally, the assumptions of the 𝑡-tests are seriously violated. In particular, if the type of
data you have is ordinal in nature and not at least interval. On such occasions an alternative
approach is to use nonparametric tests. We are not going to place much emphasis on them in
this unit as they are only occasionally used. But you should be aware of them and have some
familiarity with them.

Nonparametric tests are also referred to as distribution-free tests. These tests have the
obvious advantage of not requiring the assumption of normality or the assumption of
homogeneity of variance. They compare medians rather than means and, as a result, if the
data have one or two outliers, their influence is negated.

However, certain assumptions associated with non-parametric tests are:

(i) Sample observations are independent.

(ii) Lower order moments exist.

Advantages of non-parametric methods over parametric methods:

(i) Non-parametric methods are readily comprehensible, very simple and easy to
apply and do not require complicated sample theory.
(ii) No assumption is made about the form of the frequency function of the parent
population from which sampling is done.
(iii) Since the socio-economic data are not, in general, normally distributed, non-
parametric test have found applications in Psychometrics, Sociology and
Educational Statistics.
(iv) Non-parametric tests are available to deal with the data which are given in ranks.

Disadvantages of non-parametric methods over parametric methods:

(i) Non-parametric tests can be used only if the measurements are nominal or ordinal.
Even in that case, if a parametric test exists it is more powerful than the non-
parametric test.

1
Dr. Md. Abdus Salam Akanda Website: http://du.ac.bd
Professor of Statistics, DU E-mail: [email protected]

(ii) Non-parametric tests are designed to test statistical hypothesis only and not for
estimating the parameters.

Objectives
 Calculate Mann-Whitney Test
 Calculate Wilcoxon’s Matched-Pairs Signed-Ranks Test
 Calculate Kruskal-Wallis One-Way ANOVA
 Calculate Friedman’s Rank Test for 𝑘 Correlated Samples

Nonparametric statistics or distribution-free tests are those that do not rely on parameter
estimates or precise assumptions about the distributions of variables. In this chapter we will
learn how to use SPSS Nonparametric statistics to compare 2 independent groups, 2 paired
samples, k independent groups, and 𝑘 related samples.

2
Dr. Md. Abdus Salam Akanda Website: http://du.ac.bd
Professor of Statistics, DU E-mail: [email protected]

3
Dr. Md. Abdus Salam Akanda Website: http://du.ac.bd
Professor of Statistics, DU E-mail: [email protected]

Mann-Whitney Test
Let’s begin by comparing 2 independent groups using the Mann-Whitney Test. We want to
compare the number of stressful life events reported by cardiac patients and orthopedic
patients.
 Open stressful events.sav.
 Select Analyze/Nonparametric Tests/Two Independent Samples.
 Select data as the Test Variable and group as the Grouping Variable. Click on
Define Groups and specify 1 for Group 1 and 2 for Group 2, then click Continue.
Under Test-Type, select Mann-Whitney U. Then click on Options.
 Under Statistics, select Descriptives. Then click Continue. In the main dialog box,
click Ok.
The output follows.
Descriptive Statistics
N Mean Std. Minimu Maximu
Deviation m m
Data 11 8.6364 11.12001 .00 32.00
Group 11 1.4545 .52223 1.00 2.00

Mann-Whitney Test
Ranks
Group N Mean Rank Sum of Ranks
Cardiac Patient 6 7.50 45.00
Data Non-cardiac Patient 5 4.20 21.00
Total 11

Test Statisticsa
Data
Mann-Whitney U 6.000
Wilcoxon W 21.000
Z -1.647
Asymp. Sig. (2-tailed) .100
Exact Sig. [2*(1-tailed Sig.)] .126b

4
Dr. Md. Abdus Salam Akanda Website: http://du.ac.bd
Professor of Statistics, DU E-mail: [email protected]

a. Grouping Variable: Group

b. Not corrected for ties.

Comment: There is not a statistically significant difference in stressful life events for the 2
groups at 5% level of significance.

Wilcoxon’s Matched Pairs Signed-Ranks Test

Now, let’s compare paired or related data. We will compare the volume of the left
hippocampus in twin pairs, one of whom is schizophrenic and one of whom is normal.

To conduct a Wilcoxon signed-rank test

 Open Hippocampus Volume.sav.

1. Select the Analyze menu.
2. Click on Nonparametric Tests, then Legacy Dialogs and then on 2 Related
Samples… to open the Two-Related-Samples Tests dialogue box.
3. Select the variables you require (i.e. Normal and Schizophrenic) and click on the
button to move the variables into the Test Pair(s) List: box.
4. Ensure that the Wilcoxon check box has been selected.
5. Click on the Options… command pushbutton to open the Two-Related-Samples:
Options sub-dialogue box.
6. Ensure that the Descriptives check box has been selected.
7. Click on Continue and then OK.

The output follows.

Descriptive Statistics
N Mean Std. Minimu Maximu
Deviation m m
Normal 15 1.7593 .24150 1.25 2.08
Schizophrenic 15 1.5600 .30126 1.02 2.02

Wilcoxon Signed Ranks Test

5
Dr. Md. Abdus Salam Akanda Website: http://du.ac.bd
Professor of Statistics, DU E-mail: [email protected]

Ranks
N Mean Rank Sum of Ranks
Negative Ranks 14a 7.93 111.00
Positive Ranks 1b 9.00 9.00
Schizophrenic - Normal
Ties 0c
Total 15
a. Schizophrenic < Normal
b. Schizophrenic > Normal
c. Schizophrenic = Normal

Test Statisticsa
Schizophrenic - Normal
Z -2.897b
Asymp. Sig. (2-tailed) .004
a. Wilcoxon Signed Ranks Test
b. Based on positive ranks.

The Sum of Ranks column includes the T values. Compare them to the values in the text.
Note that the test statistic in SPSS is z. Regardless, the results are the same. There is a
significant difference in hippocampal volume between normals and schizophrenics at 5%
level of significance.

Kruskal-Wallis One-Way ANOVA

Now let’s compare more than 2 independent groups. We’ll use the example illustrated in
Table 20.4 of the text, comparing the number of problems solved correctly in one hour by
people who received a depressant, stimulant, or placebo drug.
 Open problem solving.sav.

To conduct a Kruskal-Wallis test

1. Select the Analyze menu.

2. Click on Nonparametric Tests, then Legacy Dialogs and then K Independent

6
Dr. Md. Abdus Salam Akanda Website: http://du.ac.bd
Professor of Statistics, DU E-mail: [email protected]

Samples… to open the Tests for Several Independent Samples box.

3. Select the Test variable (data) and click on the button to move the variable into the
Test Variable List: box.
4. Select the Grouping variable (group) and click on the button to move the variable into
the Grouping Variable: box.
5. Click on the Define Range… command pushbutton to open the Several Independent
Samples: Define Range sub-dialogue box.
6. Enter the first value for the independent variable in the Minimum: box, then tab. Enter
the greatest value for the independent variable in the Maximum: box. (Indicate 1 for the
Minimum and 3 for the Maximum since there are 3 groups, identified as 1,2, and 3.)
7. Click on Continue.
8. Ensure the Kruskal-Wallis H check box has been selected.
9. Click on OK.

Kruskal-Wallis Test

Ranks
Group N Mean
Rank
Cardiac Patient 7 5.00
Non-cardiac
8 14.38
Data Patient
3.00 4 10.00
Total 19

Test Statisticsa,b
Data
Chi-Square 10.407
df 2
Asymp. Sig. .005
a. Kruskal Wallis Test

7
Dr. Md. Abdus Salam Akanda Website: http://du.ac.bd
Professor of Statistics, DU E-mail: [email protected]

b. Grouping Variable: group

Comment: The result supports the conclusion that problems solved correctly varied
significantly by group at 5% level of significance.

Friedman’s Rank Test for K Related Samples

Now, let’s move on to an example with k related samples. We want to see if reading time is
effected when reading pronouns that do not fit common gender stereotypes.
 Open pronouns.sav.

To conduct a Friedman test

1. Select the Analyze menu.

2. Click on Nonparametric Tests, then Legacy Dialogs and then K Related Samples… to
open the Tests for Several Related Samples box.
3. Select the variable you require (He, She, They) and click on the button to move the
variables into the Test Variable List: box.
4. Ensure the Friedman check box has been selected.
5. Click on OK.

Friedman Test
Ranks
Mean
Rank
He 2.64
She 2.00
They 1.36

Test Statisticsa
N 11
Chi-Square 8.909
df 2

8
Dr. Md. Abdus Salam Akanda Website: http://du.ac.bd
Professor of Statistics, DU E-mail: [email protected]

Asymp. Sig. .012

a. Friedman Test

As you can see, the Chi Square value is in agreement with the one in the text. We can
conclude that reading times are related to pronoun conditions at 5% level of significance.

Exercises
1. Using birthweight.sav, use the Mann-Whitney Test to compare the birthweight of
babies born to mothers who began prenatal care in the third trimester to those who
began prenatal classes in the first trimester.
2. Using anorexia family therapy.sav, compare the subjects’ weight pre and post
intervention using Wilcoxon’s Matched Pairs Signed Ranks Test. What can you
conclude?
3. Using maternal role adaptation.sav, compare maternal role adaptation for the 3
groups of mothers using the Kruskal-Wallis ANOVA. What can you conclude?
4. Using Eysenck recall repeated.sav, examine the effect of processing condition on
recall using Friedman’s Test. What can you conclude?