> What is suggested by Doug is that when we have a block design
> experiment is more appropiate to do multiple linear regression with
> one lag?
My understanding is that usually the lag is set to 0 for block designs with 3dDeconvolve.
> So, can we talk about lags (lag 3,4 etc ) in a block design analysis
> or no?
There are some special circumstances in which you can analyze block designs with multiple lags.
Gang