Hi Didier,
I had a thorough look into your problem.
It occurs because the conditions you impose on trial selection,
0B_face_target , result in a small number of trials. For the subject
102816 in Run1 there were 8 clean trials in this conditions but the
participant response was wrong in all of them. So no trials were
selected for this condition 1 in Run1. For Condition 2 some trials were
selected. In Run 2 some trials were found for borh conditions 1 and 2.
What then happened was that the hcp_eravg pipeline does not find any
data for Condition1-Run1 so it uses only Condition1-Run2 which has 245
clean channels. But for Condition 2 it finds data both from Run1 and
Run2 , and although they have both 245 clean channels , they have only
244 common clean channels. Because this fusion of sensors was not done
for Condition 1 (as only data from Run 2 was found) there is a mismatch
of channels. During the testin we had not use conditions with small
number of trials that is why this scenario has escaped our attention.
Now I think the quickest and easiest way to proceed with your work would
be to go to function
hcp_eravg_contrasts.m
and in line 292 where you see the code:
%--------
if isempty(datacntr2)
errorFname=['NOTRIALSFOUND_',experimentid,'_',scanmnem,'_',sortcontrlist{1}{iC}.mnemprint,'_[MODE-',avgmode,'].txt'];
errorTime=clock;
errorCase=[experimentid,'_',scanmnem,'_',sortcontrlist{1}{iC}.mnemprint,'_[MODE-',avgmode,']'];
errorType=['No trials were found for condition 2'];
hcp_write_ascii(errorFname,'errorTime','errorCase','errorType');
warning(['No trial found for condition 2 in
',experimentid,' - ',sortcontrlist{1}{iC}.mnemprint]) ;
continue
end
%--------
add the "else" statement :
%--------
else
[indChA,indChB]=match_str(datacntr1.label,datacntr2.label);
datacntr1=ft_selectdata(datacntr1,'channel',indChA);
datacntr2=ft_selectdata(datacntr2,'channel',indChB);
%--------
So that the whole block would be :
%--------
if isempty(datacntr2)
errorFname=['NOTRIALSFOUND_',experimentid,'_',scanmnem,'_',sortcontrlist{1}{iC}.mnemprint,'_[MODE-',avgmode,'].txt'];
errorTime=clock;
errorCase=[experimentid,'_',scanmnem,'_',sortcontrlist{1}{iC}.mnemprint,'_[MODE-',avgmode,']'];
errorType=['No trials were found for condition 2'];
hcp_write_ascii(errorFname,'errorTime','errorCase','errorType');
warning(['No trial found for condition 2 in
',experimentid,' - ',sortcontrlist{1}{iC}.mnemprint]) ;
continue
else
[indChA,indChB]=match_str(datacntr1.label,datacntr2.label);
datacntr1=ft_selectdata(datacntr1,'channel',indChA);
datacntr2=ft_selectdata(datacntr2,'channel',indChB);
end
%--------
I have run again the case in which you received the error with this
change and it seems to work.
I hope this helped
Best
Giorgos
On 10/12/2017 12:32 PM, Jean-Didier Lemarechal wrote:
Hi,
I encountered an error when computing a contrast (difference between
conditions) for subject 102816 (it worked fine, at least without
error, for other subjects {'100307'} ; {'104012'} ; {'105923'} and
{'106521'} ;
The error comes from a mismatch in buffer dimensions : ( cf attached
picture )
Error using hcp_eravg_contrasts (line 445)
Matrix dimensions must agree.
I guess it comes from bad channels/trials difference between each
condition ?
The contrast is '0B_face_target' versus '0B_face_nontarget'
I am using megconnectome-3.0 and I attached the scripts i am using :
Main script is hcp_eravg_myContrasts_Wrkmem.m and I defined the
contrasts I use in mycontrast_Wrkmem.m which is an copy of
contrast_Wrkmem.m adapted for my needs.
Thanks in advance for your heelp !
JD
_______________________________________________
HCP-Users mailing list
HCP-Users@humanconnectome.org
http://lists.humanconnectome.org/mailman/listinfo/hcp-users
---
This email has been checked for viruses by AVG.
http://www.avg.com
_______________________________________________
HCP-Users mailing list
HCP-Users@humanconnectome.org
http://lists.humanconnectome.org/mailman/listinfo/hcp-users
