> Related to Susan's question, how you would recommend setting
> the minlag and maxlag for variable stimuli? If the videos vary from
> 2 TR to 6 TR, what would help to increase the specificity of the
> sampling of the response? Would it make sense to limit the
> sampling period to a smaller window -- say from 4-8 sec
> post-stimulus rather than 2-10 sec (what they currently have)?

The rule of thumb about minlag and maxlag is (maxlag - minlag) x TR covers the duration of the impulse response function of each event. With TR = 2 and minlag = 0, you can set up maxlag = 6 or 7. Since the specification is about the impulse response function (not the hemodynamic response function for the whole event), the length of each clip in this context is irrelevant unless you suspect the brain may respond differently with longer stimuli.

> Also, since the long stimuli are more like blocks and the short
> stimuli are more like events, is it acceptable to do two analyses --
> one for the long videos using a block design and one for the short
> videos using an event-related design? Or would that undercut
> validity since responses to the short and long videos overlap?

Sure, you can model long and short stimuli of same type separately if you want to see if they have different response. Overlapping across videos should not be a problem since it is the goal of the analysis to tear apart different causes from the scanner readout.

Gang