Which Test To Use?

[edoherty]

Short version: 2-3 IVs (date, individual, compartment), repeated-measures(compartment), 20 DVs, non-parametric.


Hello,


I have 4 subjects that were tested 16 times. However, possible changes across time or individuals are not the subjects of the experiment, so would it be appropriate to consider the number of participants 64?

During each of the 64 trials, subjects were presented with six compartments simultaneously. Numerous (about 20) different measures were taken across these six compartments.

Finally, the data is not normal, and would need a non-parametric test.

Which test(s) would be appropriate for between compartments 1-6?

My initial thinking is that a Friedman test for each variable would work: 64 rows, 6 columns. Though, I would be running 20 of them, unless there is a 20-way equivalent.


Thanks!

 

[Karabiner]

I have 4 subjects that were tested 16 times. However, possible changes across time or individuals are not the subjects of the experiment, so would it be appropriate to consider the number of participants 64?

No. The measurements within a subject are dependent, so n=4.
You could consider repeated measures ANOVA with several within-
subjects factors, even in the presence of non-normality. Don't
know how severe the deviation from normality is, though.

Or, you aggregate data e.g. aggregate a subject's 16 repeated
measurements at one date (mean, or median or whatever), and
perform Friedman.

With kind regards

K.