We are using PsPM to analyse SCRs and have reached a problem when trying to report the second level analysis.
Briefly, we have acquired GSR simultaneously during an fMRI fear learning task (@ 3T). The task goes for around 16 minutes. GSR was recorded from fingers on the left hand.
I have used the GUI to export the code required to import our raw data file (Labchart 7.2 or higher MATLAB export), develop a GLM model and first level contrasts. I then wrote a loop around this script in MATLAB to do this for all participants, as suggested in the manual. I then used the GUI to generate a second level model however, when I tried to report the second level model I got the following error message:
Subscript indices must either be real positive integers or logicals.
In file "/Applications/MATLAB_R2017b.app/toolbox/PsPM/pspm_rev2.m" (v450), function "pspm_rev2" at line 48.
In file "/Applications/MATLAB_R2017b.app/toolbox/PsPM/pspm_cfg/pspm_cfg_run_review2.m" (v450), function "pspm_cfg_run_review2" at line 10.
I am unsure at which step this problem may have been introduced.
Any advice would be greatly appreciated!
Thanks,
Hannah
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thanks for the message. We'll investigate this week. In the meantime (and actually more generally), I'd suggest exporting statistics into a more specific statistics software for the second-level analysis. This is more flexible, since the tests implemented PsPM are really just t-tests.
Best wishes
Dominik
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
To me this looks like there might be something wrong with the specification of the contrasts. But this is just a guess. Could you send me an example model file (the file you pass as an argument to the second level review method to create the second level model)? Then I can check whether there is something wrong with the model-file.
Best
Tobias
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thank you for your quick reply and helpful advice. I have attached an example file - one subject who has been imported, had a GLM created and and then calculated the first level contrasts.
In addition, we were wondering if you had any tips regarding removing MRI artefact. As far as we can understand, when constructing the GLM, selecting the the default filter settings applies a unidirectional, 0.05Hz high-pass Butterworth and 5Hz low-pass Butterworth filter. In your experience, is this sufficient to clean raw data or would you recommend removing artefacts prior to generating the model? e.g. using the median filter in Artefact Removal? When visualising our data in LabChart, it appears that applying a 0.1Hz low-pass filter provides a sufficient level of noise reduction - is this achievable in pspm?
Further, when creating the GLM, the manual recommends that if between-subject analysis it being performed, to NOT use normalisation. As per my prior enquiry, we intend to export the first level statistics and subsequently compare between subjects and conditions. Are there any examples when this would in fact be useful to apply when between-subject analysis are performed? Should we instead apply the Z-scoring when generating the first level contrasts?
Thank you for the example file. Unfortunately I am still not able to reproduce the problem you described above. Could you also send me the file you get after creating the second level contrasts (sorry, I was a bit in a rush last week and forgot to mention this)? Could you also provide me the contrasts you specify to report the second level model?
Regarding filter settings, these can be set in the GLM using 'Filter settings' > 'Edit settings' (in the GUI). For artefact removal we have automatic functions, which in the GUI are available over 'Tools' > 'Artefact removal'. These include Median and Buterworth filtering and also a simple SCR quality correction method according to http://dx.doi.org/10.1109/TBME.2017.2758643. If the automatic correction does not yield satisfying results, you could also inspect the data manually using the data editor. With the data editor you are able to mark periods of artefacts in the data. These can be saved as epoch file, which then can be specified as missing epochs in the GLM, and therefore can be excluded from the analysis. The data editor is available over 'Tools' > 'Data editor'.
Regarding normalisation, this is not an easy question. What I know is, if you normalise the data in a between subject analysis you lose the group effect and that is why we do not recommend normalising in such cases. Also z-scoring before generating first level contrasts would lead to loss of the group effect.
I hope this helps.
best
Tobias
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
We will investigate the filter settings you have described - thank you for your suggestions.
I have attached the file that is created (using our unfiltered data) after defining the second level contrasts (when setting this up - I choose for it to read the contrasts from the 1st file). When reporting the second level model I have been doing each contrast separately and in the same way that I define them for the first level contrasts:
Baseline_plus>minus [1 -1]
Conditioning_overall_plus>minus [ 0 0 0.5 -0.5 0.5 -0.5]
Conditioning_early_plus>minus [ 0 0 1 -1]
Conditioning_late_plus>minus [ 0 0 0 0 1 -1]
The programme is then unable to report these contrasts as per my original enquiry.
I have applied the artefact removal options you suggested and it does appear to clean the data well. Regardless, we are still unable to report the second level model.
Kind regards,
Hannah
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Sorry for the late response. According to the data you sent me, I think the parameter for the contrast vector is wrong. If you want to report all the contrasts, just leave it empty, or use indexes if you want to report specific contrasts (i.e. [2 4] for the 2nd and 4th contrast). Does this work? Otherwise could you also tell me what parameters you use to report the second level model?
Best
Tobias
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Dominik and Tobias,
We are using PsPM to analyse SCRs and have reached a problem when trying to report the second level analysis.
Briefly, we have acquired GSR simultaneously during an fMRI fear learning task (@ 3T). The task goes for around 16 minutes. GSR was recorded from fingers on the left hand.
I have used the GUI to export the code required to import our raw data file (Labchart 7.2 or higher MATLAB export), develop a GLM model and first level contrasts. I then wrote a loop around this script in MATLAB to do this for all participants, as suggested in the manual. I then used the GUI to generate a second level model however, when I tried to report the second level model I got the following error message:
Subscript indices must either be real positive integers or logicals. In file "/Applications/MATLAB_R2017b.app/toolbox/PsPM/pspm_rev2.m" (v450), function "pspm_rev2" at line 48. In file "/Applications/MATLAB_R2017b.app/toolbox/PsPM/pspm_cfg/pspm_cfg_run_review2.m" (v450), function "pspm_cfg_run_review2" at line 10.I am unsure at which step this problem may have been introduced.
Any advice would be greatly appreciated!
Thanks,
Hannah
Thanks for the message. We'll investigate this week. In the meantime (and actually more generally), I'd suggest exporting statistics into a more specific statistics software for the second-level analysis. This is more flexible, since the tests implemented PsPM are really just t-tests.
Best wishes
Dominik
Hi Hannah
To me this looks like there might be something wrong with the specification of the contrasts. But this is just a guess. Could you send me an example model file (the file you pass as an argument to the second level review method to create the second level model)? Then I can check whether there is something wrong with the model-file.
Best
Tobias
Hi Domink and Tobias,
Thank you for your quick reply and helpful advice. I have attached an example file - one subject who has been imported, had a GLM created and and then calculated the first level contrasts.
In addition, we were wondering if you had any tips regarding removing MRI artefact. As far as we can understand, when constructing the GLM, selecting the the default filter settings applies a unidirectional, 0.05Hz high-pass Butterworth and 5Hz low-pass Butterworth filter. In your experience, is this sufficient to clean raw data or would you recommend removing artefacts prior to generating the model? e.g. using the median filter in Artefact Removal? When visualising our data in LabChart, it appears that applying a 0.1Hz low-pass filter provides a sufficient level of noise reduction - is this achievable in pspm?
Further, when creating the GLM, the manual recommends that if between-subject analysis it being performed, to NOT use normalisation. As per my prior enquiry, we intend to export the first level statistics and subsequently compare between subjects and conditions. Are there any examples when this would in fact be useful to apply when between-subject analysis are performed? Should we instead apply the Z-scoring when generating the first level contrasts?
Thanks again,
Hannah
Hi Hannah
Thank you for the example file. Unfortunately I am still not able to reproduce the problem you described above. Could you also send me the file you get after creating the second level contrasts (sorry, I was a bit in a rush last week and forgot to mention this)? Could you also provide me the contrasts you specify to report the second level model?
Regarding filter settings, these can be set in the GLM using 'Filter settings' > 'Edit settings' (in the GUI). For artefact removal we have automatic functions, which in the GUI are available over 'Tools' > 'Artefact removal'. These include Median and Buterworth filtering and also a simple SCR quality correction method according to http://dx.doi.org/10.1109/TBME.2017.2758643. If the automatic correction does not yield satisfying results, you could also inspect the data manually using the data editor. With the data editor you are able to mark periods of artefacts in the data. These can be saved as epoch file, which then can be specified as missing epochs in the GLM, and therefore can be excluded from the analysis. The data editor is available over 'Tools' > 'Data editor'.
Regarding normalisation, this is not an easy question. What I know is, if you normalise the data in a between subject analysis you lose the group effect and that is why we do not recommend normalising in such cases. Also z-scoring before generating first level contrasts would lead to loss of the group effect.
I hope this helps.
best
Tobias
Hi Tobias,
We will investigate the filter settings you have described - thank you for your suggestions.
I have attached the file that is created (using our unfiltered data) after defining the second level contrasts (when setting this up - I choose for it to read the contrasts from the 1st file). When reporting the second level model I have been doing each contrast separately and in the same way that I define them for the first level contrasts:
Baseline_plus>minus [1 -1]
Conditioning_overall_plus>minus [ 0 0 0.5 -0.5 0.5 -0.5]
Conditioning_early_plus>minus [ 0 0 1 -1]
Conditioning_late_plus>minus [ 0 0 0 0 1 -1]
The programme is then unable to report these contrasts as per my original enquiry.
Thanks,
Hannah
Hi Domink and Tobias,
I have applied the artefact removal options you suggested and it does appear to clean the data well. Regardless, we are still unable to report the second level model.
Kind regards,
Hannah
Hi Hannah
Sorry for the late response. According to the data you sent me, I think the parameter for the contrast vector is wrong. If you want to report all the contrasts, just leave it empty, or use indexes if you want to report specific contrasts (i.e. [2 4] for the 2nd and 4th contrast). Does this work? Otherwise could you also tell me what parameters you use to report the second level model?
Best
Tobias
Hi Tobias,
It works! I now get the statistics and the graph!
Thank you very much for your help!
Hannah