Which is better Wilcoxon Signed Ranks Test or Paired Samples T-Test for my specific data?

#1
I am not sure which test to is valid: paired samples t-test of a Wilcoxon signed rank test for my data.
My data meets all the assumptions of a t-test, however, I only have 29 paired samples and the data is not normally distributed, although not terribly off. I know that once we hit 30 samples, some say it no longer matters whether the data is distributed normally. As my data is at 29 and not 'entirely' distributed normally (although I don't think terribly off), I am not sure how to proceed.
I have attached image files of what the normality tests look like or sample called Beat Heard (pics 1 and 2) and the 2nd sample called Beat No Heard (pics 3 and 4).
I also attached screen shots of when I ran the Paired Samples T-test results. The Wilcoxon was p = .031. NOTE: I am only doing one-sided for p value.
Thanks so much ahead of time for any advice you guys have! :) View attachment 3316
 

Karabiner

TS Contributor
#2
There is no attachment?!

Did you inspect your raw variables (wrong), or the differences (correct) with respect to normality?

By the way, do not perform one-sided test without very strong reason to do so.

With kind regards

Karabiner
 
#3
Ok. so after grappling with this for a week I realized I only did each variable NOT the difference. Upon running the difference the data is normal.

For the one sided, I only wanted to know if after listening to x if a certain behavior increased. that was my research question. Is it ok to proceed with one-sided then? thanks for your reply!
 

Karabiner

TS Contributor
#4
Many research questions have a direction, but you very rarely see a one-tailed test.
One simply should not make oneself voluntarily half blind.

With kind regards

Karabiner