SPSS - Mann Whitney U Test help.

#1
Hi everyone,

Just doing some stats on SPSS, which I am new to. The question I have to answer is;

Select the correct statistical technique to test the hypothesis there there is no difference in systolic blood pressure due to aerobic exercise training lasting 6, 12 and 24 weeks.

There a three groups;

Group One- Performs training for 6 weeks, Group Two- 12 weeks, Group Three - 24 weeks.

I started of by doing the Simple One Way indepedent groups ANOVA & Post Hoc Analysis, to test for normal distribution, along with Homogeneity of Variance, however the latter was violated.

Due to that I conducted a Kruskal-Wallis test, there was a significant difference, leading me to the Mann Whitney U test.

However as seen above I have three groups, and do not know how to compare all three together on Mann Whitney U test as it does not seem possible. I have ran the following Mann Whitney U tests;

Group 1-3 Group 2-3 Group 1-2

The results have come back, but they are not what you would expect. So if anyone could give advice on how to conduct the test with the 3 groups I have it would be greatly appreciated. Also when applying a bonferroni adjustment to a Mann Whitney do you apply the it to the sig?

Thanks again,
 

Karabiner

TS Contributor
#2
along with Homogeneity of Variance, however the latter was violated.
You could conduct a oneway ANOVA with Welch or Brown-Forsythe corrections
(in SPSS, see options under ONEWAY ANOVA).

By the way, if group sizes are equal, then inhomogenous variances don't matter
much with regard to comparison of means, AFAIR.

Moreover, if this was a randomized experiment (if not, then differences between
groups could not easily be attributed to different length of training), then
heteroscedascity is a result in itself! You can then already reject the
hypothesis there there is no difference in systolic blood pressure
because of the different variability between groups, due to training.


The results have come back, but they are not what you would expect.
Maybe you could share a bit more of the information you have
about the results, and in which respect they weren't like you expected.

With kind regards

K.