Hello Shayna:
I'm not sure if this is what you had in mind, but:
The residual is defined:
ei = Yi - Yhat
where
ei = ith residual
Yi = ith data point
Yhat = fitted value
The "standardized residual" is defined:
ri = ei / sqrt(MSE)
where
ri = ith standardized residual
MSE = estimate of the measurement variance
Program 3dDeconvolve can be used to calculate the single sample t-test of
population mean equal to zero, as follows:
First concatenate the datasets from each subject:
3dTcat -prefix myData Subj1+tlrc etc. etc. (do NOT use -rlt option)
Thus, dataset myData+tlrc contains one sub-brick for each subject.
Now use program 3dDeconvolve to calculate the t-test as follows:
3dDeconvolve -input myData+tlrc -mask myMask+tlrc -polort 0 \
-progress 10000 -tout -vout -bucket myTtest -errts myResid
The bucket dataset myTtest+tlrc contains 3 sub-bricks: sub-brick #0 contains
the sample mean across subjects at each voxel location; sub-brick #1 contains
the one-sample t-test values at each voxel location; and sub-brick #2 contains
the estimated error variance MSE. The dataset myResid+tlrc contains the
residual error at each voxel for each subject, where each sub-brick corresponds
to a different subject.
Now, to calculate the standardized residuals, use the 3dcalc commands:
3dcalc -a myResid+tlrc -b myTtest+tlrc'[2]' -expr "a/sqrt(b)" \
-fscale -prefix myStdRes
The output dataset myStdRes+tlrc will contain the standardized residuals
at each voxel location for each subject.
Please let me know how this works out.
Doug Ward