[color=#000000]Dear Zahra,[/color]
[color=#000000]That is perfectly fine. Generally you do not add control covariates such as QA_invalid and/or QA_MaxMotion to your second-level analysis models unless you believe there is a reasonable risk of your main effect of interest being confounding by potential residual effects of motion/outliers (e.g. if you were seeing some residual correlation between Quality-Control motion and Functional-Connectivity measures in your "distribution of QC-FC associations" Quality Assurance plots). [/color]
[color=#000000]That said, it is not a bad idea to add them there even if "just to check", as you are doing in your example. In that example, though, there are a couple of things wrong with the second-level analysis specification (hence the "WARNING!" message showing in that window). To fix that you could create two new second-level covariates containing the QA_InvalidScans and QA_MeanMotion values within the Autism group only and centered (e.g. [/color]create a new "[i]QA_InvalidScans_autism[/i]" covariate in [i]Setup.Covariates2ndLevel [/i]and enter in this covariate [i]values [/i]field the formula "[i]autism.*(QA_InvalidScans-mean(QA_InvalidScans(autism>0)))"[/i] without the quotes), and then use the same second-level analysis as you are doing now but using the QA*_autism covariates instead (you should see that the "warning" message disappears and the "n" value switches to the actual number of subjects in the autism group only)
Hope this helps
Alfonso
[i]Originally posted by zahra moradimanesh:[/i][quote]Dear all,
I have two groups (autism/control) that I had defined as 2nd level covariate in setup. Now in the 2nd level analysis, to see ROI-ROI connectivities of each group I simply select it from the Subjects Effects and click on Results Explorer, which gave me the ROI.mat file. Is this wrong? Because I wonder if I should manually include QA_invalid scans and QA_MaxMotion and set my contrast to something like [1 0 0] in order to exclude invalid scans and scans with large motion before running the analysis? But on the other hand I suppose these corrections has already been done during the preprocessing step, so maybe there is no need to manually include them in the design matrix of my 2nd level analysis, and QA_ covariates are just there as reports?
Sorry for my confusion.
And thanks a lot![/quote]
[color=#000000]That is perfectly fine. Generally you do not add control covariates such as QA_invalid and/or QA_MaxMotion to your second-level analysis models unless you believe there is a reasonable risk of your main effect of interest being confounding by potential residual effects of motion/outliers (e.g. if you were seeing some residual correlation between Quality-Control motion and Functional-Connectivity measures in your "distribution of QC-FC associations" Quality Assurance plots). [/color]
[color=#000000]That said, it is not a bad idea to add them there even if "just to check", as you are doing in your example. In that example, though, there are a couple of things wrong with the second-level analysis specification (hence the "WARNING!" message showing in that window). To fix that you could create two new second-level covariates containing the QA_InvalidScans and QA_MeanMotion values within the Autism group only and centered (e.g. [/color]create a new "[i]QA_InvalidScans_autism[/i]" covariate in [i]Setup.Covariates2ndLevel [/i]and enter in this covariate [i]values [/i]field the formula "[i]autism.*(QA_InvalidScans-mean(QA_InvalidScans(autism>0)))"[/i] without the quotes), and then use the same second-level analysis as you are doing now but using the QA*_autism covariates instead (you should see that the "warning" message disappears and the "n" value switches to the actual number of subjects in the autism group only)
Hope this helps
Alfonso
[i]Originally posted by zahra moradimanesh:[/i][quote]Dear all,
I have two groups (autism/control) that I had defined as 2nd level covariate in setup. Now in the 2nd level analysis, to see ROI-ROI connectivities of each group I simply select it from the Subjects Effects and click on Results Explorer, which gave me the ROI.mat file. Is this wrong? Because I wonder if I should manually include QA_invalid scans and QA_MaxMotion and set my contrast to something like [1 0 0] in order to exclude invalid scans and scans with large motion before running the analysis? But on the other hand I suppose these corrections has already been done during the preprocessing step, so maybe there is no need to manually include them in the design matrix of my 2nd level analysis, and QA_ covariates are just there as reports?
Sorry for my confusion.
And thanks a lot![/quote]