Hi,
I have a problem using 3dMEMA on surface data.
I have done all the analysis on standard mesh with ld=60. On individual subject, 3dDevonvolve was done to get coef and Tstat for each regressor and the -bucket option was used to save the data. When viewing on individual subject in suma, the result looks very reasonable. I set the AFNI_NIML_TEXT_DATA environment parameter to YES just so that I can also check result in Matlab.

And then I used 3dcopy to copy the "bucket" niml.dset files from different subjects to the same folder to do group analysis. The 3dMEMA command is like this:

3dMEMA -prefix filled.common.groupMEMASurf.$hemi -max_zeros 2 \
-missing_data 0 -model_outliers -set common_resp 02 filled.stats.02.$hemi.niml.dset'[3]' filled.stats.02.$hemi.niml.dset'[4]' \
03 filled.stats.03.$hemi.niml.dset'[3]' filled.stats.03.$hemi.niml.dset'[4]' \
06 filled.stats.06.$hemi.niml.dset'[3]' filled.stats.06.$hemi.niml.dset'[4]' \
07 filled.stats.07.$hemi.niml.dset'[3]' filled.stats.07.$hemi.niml.dset'[4]' \
09 filled.stats.09.$hemi.niml.dset'[3]' filled.stats.09.$hemi.niml.dset'[4]'
$hemi was either lh or rh.
It did run through, but the resulting data was strange. It gave me a .1D file and +orig.HEAD and +orig.HEAD file. Apparently I cannot view the +orig files. I could load the 1D file into suma and there were 4 sub-bricks, the same as what 3dMEMA outputs for volume space data. The activation pattern looks correct, but the statistic parameter (p-value) associated with the t-stat is missing. Also the labels of the sub-bricks are all called "numeric". Reading into the 1D file, it does have ni_stat to label the degree of freedom for the d-stat but this seems to be not read into suma.

I am wondering what would be the correct way to use 3dMEMA for surface data. I tried to convert the "bucket" niml file to binary niml format before processing. But it still comes out the same way.

Thanks!
Mingbo