a single trial data analysis ("trial by trial")

Hi, List experts,

I am attempting to do a single trial data analysis - some others call it trial-by-trial analysis. I know that in the 3dDeconvolve, we simply use -stim_times_IM for regressors of interest. However, for each event, we then have a lot of trials (i.e., data), then how should I do the 3dANOVA, or 3dttest?

For instance, in my experiment, there are three trial conditions: Post_emotion, neutral, Neg_emotion. Then each condition, I have almost 150 trials after I run the 3dDeconvolve (with -stim_times_IM). Does it mean that I just extract the mean data from those trials, and then run the group analysis (e.g, 3dANOVA, or 3dttest)? Any suggestions on this analysis would be highly appreciated.


Hi Juan,

Exactly why are you using the _IM form of that,
i.e. why are you using -stim_times_IM instead
of simply -stim_times? It sounds like perhaps
you do not need one beta per event, in which
case using the simple -stim_times is probably
more appropriate.

  • rick

Hi, rick,

Thanks for letting me know your comments. I found that some people do the whole brain network analysis, also the MVPA, using the -stim_times_IM in the 3dDeconvolve. I tried it and found that by this way,we can produce a lot of data trials. This reminds me that some people claimed that a single trial data analysis (or, trial-by-trial analysis) is very powerful.

Clearly, this is different from the common -stim_times. Do you have any experience with a single trial data analysis (“trial by trial”), and can introduce some? Am I wrong on something?


I have not done this and was not certain you really meant
to use IM given the notion of an ANOVA or t-test. But
ignoring those uses, one can certainly imagine doing
something like a correlation analysis. Surely there are
many similar options.

  • rick