Hi Michael,
A couple of things come to mind. But nothing too serious. As long as you estimate the smoothness of the data that is used to generate the statistics you are OK.
Also, blurring time series, or beta coefficients is OK. Blurring F and T stats is less so. So if you were to conduct 2nd level analysis using both beta and variance estimates like with 3dMEMA, then you are better off blurring the time series before you do the 1st level (single subject) regression analysis. This way you do not have to worry about normalizing statistics before blurring them.
cheers,
ziad