Hi there

I had a query about scaling the beta coefficients to percent signal change which then lead to checking my script to make sure I have all the nuts and bolts for any fMRI analysis. I have attached parts of the script I use. Can you clarify on the stance of normalisation when one stitches together multiple runs? what is best way to scale or normalise? Is it better to run the analysis and then scale the betas at the very end? Or is there another way?

Also, how does a stand alone 3dREMLFit function (like the one in my script below) work as far as scaling is concerned? Does it take care of scaling without a flag?Can the output from this 3dREMLFit code be used to extract beta coefs without worrying about scaling them?

Thank you.


afni_proc.py -subj_id $i \
-dsets $run1 $run2 $run3 $run4 $run5 \
-do_block align tlrc \
-copy_anat $t1 \
-volreg_align_to last \
-volreg_align_e2a \
-volreg_tlrc_warp \
-blur_in_automask \
-blur_size 8 \
-mask_apply anat \
-regress_stim_files $design1 $design2 $design3 $design4 $design5 $design6 $design7 $design8 $design9 $design10 \
-regress_stim_labels $name1 $name2 $name3 $name4 $name5 $name6 $name7 $name8 $name9 $name10 \
-regress_basis 'SPMG1' \
-regress_censor_motion 1.0 \
-regress_opts_3dD -rout \
-regress_censor_first_trs 2 \
-regress_opts_3dD -rout \
-regress_est_blur_epits \
-regress_est_blur_errts \
-regress_use_stim_files


3dREMLfit -matrix X.xmat.1D \
-input "pb04.${i}.r01.scale+tlrc.HEAD pb04.${i}.r02.scale+tlrc.HEAD pb04.${i}.r03.scale+tlrc.HEAD pb04.${i}.r04.scale+tlrc.HEAD pb04.${i}.r05.scale+tlrc.HEAD" \
-fout -rout -tout -Rbuck stats.${i}_REML -Rvar stats.${i}_REMLvar \
-Rfitts fitts.${i}_REML -Rerrts errts.${i}_REML -verb