Performing SPM using individual trials or the trial mean as input?

[Vincent Fohanno]

Hi,

I am performing a comparison between two methods using the spm1d package.
I have around 15 subjects and want to perform a spm analysis (paired t-tests) on the joint angles and moments.
My question is: should I input all trials or the trial means (mean of all trials per subject)?
I tried both and get very different output in terms of statistical differences, more specifically I found way more differences when using each trial as input instead of the trial mean.

Thank you for you guidance on this topic.

Vincent

[Ruhollah Basatnia]

In my opinion, the main question should be: Is studying interpersonal variation important to you or not?
Depending on your research questions, you should perform one of the analyses you mentioned. If the variability of the data among the subjects is high or the type of your intervention has increased the variability of the data, you should use the first type of analysis (all trials). But if all subjects' changes are very similar, you may be able to use the mean of the data with caution (the trial means).

[Lauri Stenroth]

If you use all trials in the analysis you will have both within and between participant data entering the statistical analysis which would require a mixed model for the analysis between conditions (I'm assuming that you have different conditions between which you want to assess the effect, e.g., pre vs post intervention). If there is no option to perform mixed modeling in SPM you should use the average of the trials within the participant and condition as the input. As said in the previous response, here the interest is not in the within participant and condition variability but in the effect between conditions.

[Vincent Fohanno]

Thanks a lot to both of you.