Hi Michael,

Thanks for your clarification. I understand the experiment design better now. So, the three categories of stimulus temperature (hot, warm, and cold) are based on each subject's perception, and thus all subject don't necessarily receive a condition (e.g. warm) at the same temperature, right?

Still if each stimulus file corresponds to one specific temperature, it seems to me that scaling/weighting the stimulus timing is not justified. The purpose of such a file is to tell the program when and whether a stimulus is present (1) or absent (0), and the timing does not carry any more meanings (e.g. relative magnitude of the stimulus) than just timing. You can make the timing file look like some ratios, but in the end regression analysis would give you proportional adjustment on the beta coefficients with the corresponding statistics (t or F) remaining the same since scaling does not change its variance.

In theory, base on what each subjected reported as hot, warm, and cold, what you should find are activated areas that match what is reported as cold, warm, hot. Partial Fs values would indicate the hot stimuli having the highest partial F value, the warm stimuli having the middle F value, and the cold stimuli should having the lowest partial F value. However, a lot a times that isn't the case, sometimes all 3 levels have the same intensity, some times hot and warm response are around the same, with cold response being having the lowest instensity.

I understand your hypothesis (activation is expected to have the same trend as the stimulus temperature), but I can't see anything obviously missing in the regression analysis. Maybe there is some covariate? Or some hiding factors? Hope other people chime in and provide some thoughts about this.

Gang