PPI confusion

AFNI Message Board
1

Hello!

I am working my way through a PPI analysis and it seems that there are differing opinions on the definition of what should compose the seed timeseries. At the bootcamp I believe it was mentioned (and it’s in the powerpoint) that it should be the error time series of the ROI, but I’m having trouble finding support for this online. The powerpoint also mentions it’s something that is supposed to be “above and beyond” the main task, could someone elaborate on this idea?

Thank you,
B. Taylor Hilton

2

I work with Taylor and was similarly confused by this…

I thought one would extract the seed from the original time-series, but he said we should use the error time series (and he just returned from a great AFNI workshop). Is this the error time-series in a 3ddeconvolve call without .1D stimulus time files (from a task)?

Gang’s website is very helpful, but it just says the “average time series of the ROI”. I know that seed gets 3dDetrend-ed, etc. etc., so that feels odd to do to an error time-series. Any help or thoughts are greatly appreciated!

Thanks much!

3

It’s probably better to use a different word than “error” in the phrase “error time series”. Since the PPI analysis focuses on the fluctuations relative to the average response (or main effect) across the trials, with the approach laid out in Rick’s powerpoint slides you remove the average effect first through regression, and what is left (i.e., "error time series) would basically reflect the fluctuations across trials.

The webpage of mine you’re referring to adopts a slightly different approach in the sense that the average effect is not removed when creating the seed regressor, but in the end the two approaches are virtually identical.

4

Thanks Gang,

A related but more specific follow-up, I’m trying to run the final 3dDeconvolve command with all of our runs but it gives me a warning that the seed timeseries is too long for the given input (since each run is 552 timepoints and the total is 2208 for the seed timeseries). I imagine this could cause an issue if it isn’t reading it in correctly, any ideas?

The specific warning message:
“-stim_file 11: file length is 2208, longer than expected 552 (from dataset)”

  • Taylor
5

Taylor, do you mind posting the whole 3dDeconvolve script?

“-stim_file 11: file length is 2208, longer than expected 552 (from dataset)”

So, the seed time series is created by combining all the 4 runs? Did you also feed all the whole-brain data from the four runs as input for 3dDeconvolve?

6

Here’s the 3dDeconvolve portion. And yes the seed time series is the 4 runs combined, the four runs are put into 3dDeconvolve individually which might be the issue?

seed=PPI_${roi}/Seed_ts.1D
bp=PPI_${roi}/inter_neur_bp.1D
lp=PPI_${roi}/inter_neur_lp.1D
br=PPI_${roi}/inter_neur_br.1D
lr=PPI_${roi}/inter_neur_lr.1D
concat=PPI_${roi}/concat.1D

cd …

3dDeconvolve -input ${1}_cards_order1_dtvdFr_06mm_perc+orig ${1}_cards_order2_dtvdFr_06mm_perc+orig ${1}_cards_order3_dtvdFr_06mm_perc+orig ${1}_cards_order4_dtvdFr_06mm_perc+orig -nfirst 0 -polort A
-num_stimts 15 -basis_normall 1
-censor ${1}_cards_CENSOR.1D
-local_times
-stim_times 1 ~/Volumes/Hanson/Pitt_PYS/proc/Timing_Files/Cards/matlab_files/${1}_card_behavioral_data/Combined_Timing/bp.1D ‘GAM(8.6,.547)’
-stim_label 1 BIG_PUNISHMENT
-stim_times 2 ~/Volumes/Hanson/Pitt_PYS/proc/Timing_Files/Cards/matlab_files/${1}_card_behavioral_data/Combined_Timing/br.1D ‘GAM(8.6,.547)’
-stim_label 2 BIG_REWARD
-stim_times 3 ~/Volumes/Hanson/Pitt_PYS/proc/Timing_Files/Cards/matlab_files/${1}_card_behavioral_data/Combined_Timing/lp.1D ‘GAM(8.6,.547)’
-stim_label 3 SMALL_PUNISHMENT
-stim_times 4 ~/Volumes/Hanson/Pitt_PYS/proc/Timing_Files/Cards/matlab_files/${1}_card_behavioral_data/Combined_Timing/lr.1D ‘GAM(8.6,.547)’
-stim_label 4 SMALL_REWARD
-stim_file 5 ${1}_cards_motion.txt[1]
-stim_label 5 roll
-stim_base 5
-stim_maxlag 5 1
-stim_file 6 ${1}_cards_motion.txt[2]
-stim_label 6 pitch
-stim_base 6
-stim_maxlag 6 1
-stim_file 7 ${1}_cards_motion.txt[3]
-stim_label 7 yaw
-stim_base 7
-stim_maxlag 7 1
-stim_file 8 ${1}_cards_motion.txt[4]
-stim_label 8 I_S
-stim_base 8
-stim_maxlag 8 1
-stim_file 9 ${1}_cards_motion.txt[5]
-stim_label 9 R_L
-stim_base 9
-stim_maxlag 9 1
-stim_file 10 ${1}cards_motion.txt[6]
-stim_label 10 A_P
-stim_base 10
-stim_maxlag 10 1
-stim_file 11 $seed
-stim_label 11 seed
-stim_file 12 $bp
-stim_label 12 inter_bp
-stim_file 13 $lp
-stim_label 13 inter_lp
-stim_file 14 $br
-stim_label 14 inter_br
-stim_file 15 $lr
-stim_label 15 inter_lr
-num_glt 5
-glt_label 1 r_p
-gltsym ‘SYM: +.5BIG_REWARD +.5SMALL_REWARD -.5SMALL_PUNISHMENT -.5BIG_PUNISHMENT’
-glt_label 2 Br_Bp
-gltsym ‘SYM: +BIG_REWARD -BIG_PUNISHMENT’
-glt_label 3 sr_sp
-gltsym ‘SYM: +SMALL_REWARD -SMALL_PUNISHMENT’
-glt_label 4 Br_Sr
-gltsym ‘SYM: +.5BIG_REWARD +.5SMALL_REWARD’
-glt_label 5 Bp_Sp
-gltsym ‘SYM: +.5BIG_PUNISHMENT +.5SMALL_PUNISHMENT’
-xjpeg PPI${roi}/${1}glm_matrix_MultBlocks.jpg -rout -tout -bucket PPI${roi}/PPI_GAM_def_noNRT

7

Okay looks like it was a couple of errors in the preprocessing not related to the 3dDeconvolve step. Appreciate the help!