Comparing differences between groups in different samples - which statistical test to chose?

#1
Hello everybody

I am doing some research in sleep medicine where I compare the test results of different sleep and vigilance tests. Within these different vigilance test results, the data is sorted in different patient groups. Since my data is non-normally distributed, I performed a Kruskal-Wallis test in order to assess whether there were significant differences among the values obtained for each diagnostic group in each sleep test. Then I performed a Post Hoc test in order to assess between which diagnostic groups these differences where. In order to show which differences between which groups were more important, I performed an Effect Size calculation on each group comparison.

Now I would like to show, e.g. that the difference between patient groups A and B, is greater in Sleep Test X than in Sleep Test Y. I suspect that simply comparing the Effect Size results across Sleep Test samples would be incorrect. Does somebody know how I can correctly perform this task?

My goal is to show that certain diagnostic patient groups are more easily differentiated from others in certain Sleep Tests than in others.

Thank you very much for your support!
 

Karabiner

TS Contributor
#2
This is a repeated measured design, with an additional grouping factor.
The interaction between the repeated-measures factor (test type) and
group is what you are looking for. You will probably have to standardize
both measurements, maybe by using z-standardization.

With kind regards

Karabiner
 
#3
This is a repeated measured design, with an additional grouping factor.
The interaction between the repeated-measures factor (test type) and
group is what you are looking for. You will probably have to standardize
both measurements, maybe by using z-standardization.

With kind regards

Karabiner
Dear Karabiner
Thank you very much for your reply. Does the repeated measure design still apply if the values of each sleep test do not necessarily belong to the same patients? And what would be the additional grouping factor?

Thank you very much
redhulk21
 

Karabiner

TS Contributor
#4
Apparently I misunderstood your design. I thought that each group was given all tests, but that was not the case?

With kind regards

Karabiner
 
Last edited:
#5
Not all patients underwent all tests necessarily. And since the tests do not all test the same ability, they all use different units and thus I cannot compare them directly. I attached an example of the data I am dealing with.

Would a MANOVA be appropriate? Or should I standardise the data and then perform an ANOVA for repeated measures?

Thank you for your help!
redhulk21
 

Attachments