> can we use the –stim_times and –iresp to output IRF for irregular
> timing stimuli?

Yes, I believe this is the very reason Bob implemented the option of -stim_times.

> I am confused about this, because using -stim_times option for
> 3dDeconvolve, you have to chose one of eight rtype, however, if
> you use –iresp option to output IRF, the 3dDeconvolve supposedly
> will do deconvolution instead of multiple regression, which means
> you do NOT need to input response model.

The modeling with irregular stimulus timing is done through multiple basis functions such as those lised for option -rtype. A coefficient (beta weight) is obtained through regression analysis for each basis function of each stimulus. However each coefficient only corresponds to one basis function for a stimulus. The user usually is not interested in those individual basis function coeficients, but wants to know the whole response function. You are correct that -iresp usually is used to obtain impulse response functions for deconvolution analysis, but if -iresp option is bundled with -stim_times, the new version of 3dDeconvolve reassembles a hemodynamic response function (or impulse response function in the case of event-related designs) for each stimulus (linear combination of those basis functions).

Gang