How to do an unpaired t-test?

I have data that changes gradually across 120 trials. I would like to do an unpaired t-test to see if the first 5 trials are significantly different from the last 5 trials across all my subjects. Should I use ttest2 for this?

[Star Strider]

Another option for unpaired tests is the ranksum function.

Answer 1

[Jeff Miller]

I'm guessing you have some kind of numerical score in each of 120 trials for each of (say) 30 subjects and you want to know if the scores tend to be higher vs lower in the first 5 trials than in the last 5 trials.

If that's right, then the answer is no, don't use ttest2. Instead, compute two averages for each subject--one average of the first 5 trials and one average of the last 5 trials. Then, use ttest to do a paired t-test comparing the first-5 and last-5 averages.