3dDeconvolve dies

AFNI Message Board
,
1

Hi,experts~
I was running preprocessing on set of event-related data and there was something wrong with 3dDenconlve.My data had 3 runs, and each run scanned 202 times(TR=2s). But my stimuli was 1.5s once,and the the file in regress_stim_times had 648 events.I know the the number of time points should more than parameters,but my event time file(648) was more than scan times(606). What should I do to solve this problem and get the beta value for each event?
Below were the error and codes.Looking forward to your reply.

Error:

++ Input polort=3; Longest run=404.0 s; Recommended minimum polort=3 ++ OK ++
++ -stim_times using TR=2 s for stimulus timing conversion
++ -stim_times using TR=2 s for any -iresp output datasets
++  [you can alter the -iresp TR via the -TR_times option]
++ ** -stim_times NOTE ** guessing GLOBAL times if 1 time per line; LOCAL otherwise
++ ** GUESSED ** -stim_times_IM 1 using LOCAL times
**++ Number of time points: 606 (no censoring)**
** + Number of parameters:  678 [30 baseline ; 648 signal]**
**** ERROR: Regression model has too many parameters for dataset length :(**
**** FATAL ERROR: 3dDeconvolve dies: Insufficient data (606) for estimating 678 parameters**
** Program compile date = Aug  8 2023

Codes:
afni_proc.py
-subj_id {s}_MVPA_MNI \ -dsets {wf}/S${s}/{session}/event_run?.nii.gz \ -blocks tshift align volreg scale regress \ -copy_anat {wf}/S${s}/{session}/results/SSwarper_MNI/anatSS.{s}.nii
-volreg_align_e2a
-regress_motion_per_run
-regress_stim_types IM
-regress_censor_motion 0.4
-regress_stim_times {wf}/MVPA/seq_wholeMVPA/sub{s}/${session}/whole_time.txt
-regress_stim_labels trials
-regress_basis 'BLOCK(1.5,1)'
-regress_make_cbucket yes
-regress_opts_3dD
-jobs 2
-execute

event time(whole_time.txt):
|20|21.5|23|24.5|26|27.5|29|30.5|32|33.5|35|36.5|38|39.5|41|42.5|44|45.5|47|48.5|50|51.5|53|54.5|56|57.5|59|60.5|62|63.5|65|66.5|68|69.5|71|72.5|84|85.5|87|88.5|90|91.5|93|94.5|96|97.5|99|100.5|102|103.5|105|106.5|108|109.5|111|112.5|114|115.5|117|118.5|120|121.5|123|124.5|126|127.5|129|130.5|132|133.5|135|136.5|146|147.5|149|150.5|152|153.5|155|156.5|158|159.5|161|162.5|164|165.5|167|168.5|170|171.5|173|174.5|176|177.5|179|180.5|182|183.5|185|186.5|188|189.5|191|192.5|194|195.5|197|198.5|208|209.5|211|212.5|214|215.5|217|218.5|220|221.5|223|224.5|226|227.5|229|230.5|232|233.5|235|236.5|238|239.5|241|242.5|244|245.5|247|248.5|250|251.5|253|254.5|256|257.5|259|260.5|274|275.5|277|278.5|280|281.5|283|284.5|286|287.5|289|290.5|292|293.5|295|296.5|298|299.5|301|302.5|304|305.5|307|308.5|310|311.5|313|314.5|316|317.5|319|320.5|322|323.5|325|326.5|338|339.5|341|342.5|344|345.5|347|348.5|350|351.5|353|354.5|356|357.5|359|360.5|362|363.5|365|366.5|368|369.5|371|372.5|374|375.5|377|378.5|380|381.5|383|384.5|386|387.5|389|390.5|
|16|17.5|19|20.5|22|23.5|25|26.5|28|29.5|31|32.5|34|35.5|37|38.5|40|41.5|43|44.5|46|47.5|49|50.5|52|53.5|55|56.5|58|59.5|61|62.5|64|65.5|67|68.5|82|83.5|85|86.5|88|89.5|91|92.5|94|95.5|97|98.5|100|101.5|103|104.5|106|107.5|109|110.5|112|113.5|115|116.5|118|119.5|121|122.5|124|125.5|127|128.5|130|131.5|133|134.5|146|147.5|149|150.5|152|153.5|155|156.5|158|159.5|161|162.5|164|165.5|167|168.5|170|171.5|173|174.5|176|177.5|179|180.5|182|183.5|185|186.5|188|189.5|191|192.5|194|195.5|197|198.5|210|211.5|213|214.5|216|217.5|219|220.5|222|223.5|225|226.5|228|229.5|231|232.5|234|235.5|237|238.5|240|241.5|243|244.5|246|247.5|249|250.5|252|253.5|255|256.5|258|259.5|261|262.5|272|273.5|275|276.5|278|279.5|281|282.5|284|285.5|287|288.5|290|291.5|293|294.5|296|297.5|299|300.5|302|303.5|305|306.5|308|309.5|311|312.5|314|315.5|317|318.5|320|321.5|323|324.5|338|339.5|341|342.5|344|345.5|347|348.5|350|351.5|353|354.5|356|357.5|359|360.5|362|363.5|365|366.5|368|369.5|371|372.5|374|375.5|377|378.5|380|381.5|383|384.5|386|387.5|389|390.5|
|16|17.5|19|20.5|22|23.5|25|26.5|28|29.5|31|32.5|34|35.5|37|38.5|40|41.5|43|44.5|46|47.5|49|50.5|52|53.5|55|56.5|58|59.5|61|62.5|64|65.5|67|68.5|82|83.5|85|86.5|88|89.5|91|92.5|94|95.5|97|98.5|100|101.5|103|104.5|106|107.5|109|110.5|112|113.5|115|116.5|118|119.5|121|122.5|124|125.5|127|128.5|130|131.5|133|134.5|146|147.5|149|150.5|152|153.5|155|156.5|158|159.5|161|162.5|164|165.5|167|168.5|170|171.5|173|174.5|176|177.5|179|180.5|182|183.5|185|186.5|188|189.5|191|192.5|194|195.5|197|198.5|212|213.5|215|216.5|218|219.5|221|222.5|224|225.5|227|228.5|230|231.5|233|234.5|236|237.5|239|240.5|242|243.5|245|246.5|248|249.5|251|252.5|254|255.5|257|258.5|260|261.5|263|264.5|276|277.5|279|280.5|282|283.5|285|286.5|288|289.5|291|292.5|294|295.5|297|298.5|300|301.5|303|304.5|306|307.5|309|310.5|312|313.5|315|316.5|318|319.5|321|322.5|324|325.5|327|328.5|338|339.5|341|342.5|344|345.5|347|348.5|350|351.5|353|354.5|356|357.5|359|360.5|362|363.5|365|366.5|368|369.5|371|372.5|374|375.5|377|378.5|380|381.5|383|384.5|386|387.5|389|390.5|

2

Hi Jane,

A small initial point, it is good to include blocks of code in message board posts in 2 lines of ``` triple quotes to block the odd formatting.

But indeed, that is a long list of events. Consider this part of the output:

**++ Number of time points: 606 (no censoring)**
** + Number of parameters:  678 [30 baseline ; 648 signal]**

Since this is running IM with only 1 stimulus condition, there seem to 648 events, but only 606 time points to estimate the responses from. Notably, most of the event pairs are only 1.5s apart (TR=2s), which would probably make estimation of the response magnitudes mostly noise, even if they could be estimated. But in any case, this regression is not possible because there are more regressors (including 30 baseline ones) than time points. This experimental design is not suited for IM regression, unfortunately.

  • rick
3

Thank you for your suggestions!
Oh no... I forgot to tell you that there were 3 stimulus conditions in 648 events. I want to know if it was appropriate to divide the whole_time.txt file into 3 files according to the type of stimulus,and calculate the beta value of each event.
Jane

4

Hi Jane,

Either way, 648 events is far too many. Starting with 606 time points and 30 baseline regressors, the model can have a maximum of 576 regressors of interest, at which point the output would be almost complete garbage.

Using 3dLSS might be a possibility, but with events more frequent than the TR, it is not clear what could be gleaned from the data.

  • rick