PET data analysis requires some steps that aren't in the usual AFNI sequence. I'm not really familiar with PET data or its analyses. The one thing that I know is usually done is a global normalization of each image so that the spatial average number of counts is the same -- this is to account for dose differences between images. This could be done with a combination of 3dmaskave (with no mask) and 3dcalc. After that, I suppose 3dDeconvolve with minlag=maxlag=0 could be used to test various contrasts, using properly designed regressors. You probably would want to spatially blur (3dmerge -1blur_fwhm) the PET images before such an analysis.
Zorkon the nonPETolitrous