> From [log in to unmask] Wed Dec 1 18:50:17 1999
> Date: Wed, 1 Dec 1999 13:49:46 -0500
Dear Yanjun,
> Thank you for your time. I have 22 scans ( 11 subjects, 2 conditions,
> 1scan/condition) of PET images. I performed a paired t test in
> PET/SPECT models. The options are as follow:
>
> Population main effect: 2 cond's, 1scan/cond (paired t test)
>
> #subjuect's 11
>
> proprotional scaling
>
> PropSca global mean to 50
>
> Threshold masking prop'nal
>
> analysis thres (prop'n of global) 0.8
>
> Global calculation : mean voxel value (within per image fullmean/8
> mask)
>
> When I want to analyse the results, I use a contrast { 1 -1} or {-1 1},
> masked with no other contrasts. But if I use a corrected height
> threshold (0.05), it will correspond to very high T threshold and no
> clusters survived. If I use non-corrected height threshold (0.01 or
> 0.005), there will be some clusters survived. I'm just wondering if
> anything is wrong in my analysis process. Thank you very much.
Your analysis appears fine. The design probably has low sensitivity
because you have low degrees of freedom (with only 2 scans per
subject). You can use the uncorrected p values (or small volume
corrected p values [S.V.C. in results]) if you had predicted any
regions in advance.
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|