Unpaired or paired test?

#1
Hi,
I have been doing the following experiment:
I have recorded the startle amplitudes triggered by two different sounds - this is for one nice little rat called Roger. The sounds (12x S1 and 12x S2) were randommly distributed like this:
S1 - S1 - S1 - S2 - S2 - S1 - S2 - S2 - S2 - S1 - S2 - S1 - S1 - S2 - S1 - S1 - S2 etc...up to 24 sounds in total.

I want to check if there is a difference between the startle amplitudes triggered by S1 and the ones triggered by S2.

Both population do not follow normality hence non-parametric test.
Because the data are from the same animal, I did the Wilcoxon signed-rank test (paired test).

Am I right? If not, could you explain why? Many thanks
Julie
 

SiBorg

New Member
#2
Hi there

I don't think a paired test is correct in this instance (for a single rat) because the S1 and S2 measurements are not logically paired in your experimental design.

On the other hand, if you had 10 rats and had compared S1 and S2 for all 10 rats, then you'd pair the results for each rat. I'm not sure of the best way to pair multiple measurements though.