Hi everyone,
We collected multi-echo data on an N-back task. We preprocessed the data using the meica.py script, and then performed the first level analyses on the denoised timeseries. However, when we extract the betas from load-related regions, we see mostly negative betas, which is not what we predicted. One possibility is that we are somehow overfitting our baseline during 3ddeconvolve. Here’s my question:
Given a dataset that has been denoised with a multi-echo ICA, what polort would you recommend using at the first level analysis?
Here’s our 3ddeconvolve script for reference.
3dDeconvolve \
-input ${subject}.run2.scale_al+orig.HEAD ${subject}.run3.scale_al+orig.HEAD \
-censor ${stmdir}/${subject}.task_censor.1D \
-ortvec ${stmdir}/${subject}.task.motion.1D motion \
-polort A \
-num_stimts 7 \
-stim_times 1 ${stmdir}/${subject}.shock.stimtimes.1D 'BLOCK(1,1)' \
-stim_times 2 ${stmdir}/${subject}.button.stimtimes.1D 'BLOCK(1,1)' \
-stim_times 3 ${stmdir}/${subject}.instruction.stimtimes.1D 'BLOCK(8,1)' \
-stim_label 1 "shock" \
-stim_label 2 "button" \
-stim_label 3 "instructions" \
\
-stim_times 4 ${stmdir}/${subject}.safeone.stimtimes.1D 'BLOCK(42,1)' \
-stim_times 5 ${stmdir}/${subject}.safethree.stimtimes.1D 'BLOCK(42,1)' \
-stim_times 6 ${stmdir}/${subject}.thrtone.stimtimes.1D 'BLOCK(42,1)' \
-stim_times 7 ${stmdir}/${subject}.thrtthree.stimtimes.1D 'BLOCK(42,1)' \
-stim_label 4 "safeone" \
-stim_label 5 "safethree" \
-stim_label 6 "thrtone" \
-stim_label 7 "thrtthree" \
-tshift \
-bout \
-fout \
-rout \
-tout \
-vout \
-x1D X.xmat.1D \
-xjpeg X.jpg \
-x1D_uncensored X.nocensor.xmat.1D \
-fitts ${subject}.task.fitts \
-errts ${subject}.task.residuals \
-cbucket ${subject}.task.cbuck \
-bucket ${subject}.task.buck \
-jobs 16