help
help > RE: second-level analysis in Conn
Aug 19, 2015 01:08 AM | Alfonso Nieto-Castanon - Boston University
RE: second-level analysis in Conn
Hi Diana,
Your previous post seems to contain only one attachment (named DifferencesContrasts_ASD1_ASD2_ASD3_Social_Interpersonal.png and containing a figure labeled "connectivity between SPL l and tolTG l"). The NITRC forum only allows you to attach a single file to a post, could you please either zip your multiple screenshots into a single compressed file and attach that one, or send me by mail the original screenshot files?
In the meantime, regarding your other questions:
1) yes, you may simply select all of the ROIs typically associated with that network and either enter a between-sources contrast "eye(N)" (where N is the number of ROIs; e.g. [1 0 0;0 1 0;0 0 1] for three ROIs), which will test your effect across any of the regions in the network instead of on a single region, or more typically enter a between-sources contrast "ones(1,N)/N" (e.g. [1/3 1/3 1/3] for three ROIs) which will test your effect when looking at the average connectivity with all of the regions in the network instead of a single region. note: the between-subject effects / contrasts would be just the same as before, you would only be modifying the selected sources and the associated between-sources contrast
2) I am not entirely clear if this is what you mean here, but you can right-click on any results table (or ctrl-click if on a Mac) and export that as a text file, as a .csv spreadsheet, or as a .mat Matlab file. If, instead, you mean modifying the barplot displays, you could type the following in Matlab command-line "set(gcf,'menubar','figure')" (without the double-quotes) and then use the "Edit" options there to modify these displays, or you could use the "import values" button on the main CONN gui to get the individual connectivity values for each subject if you prefer to analyze those separately and create your own displays, for example.
3) The error bars are 90% confidence intervals for the parameter estimates (those are typically larger than standard errors and smaller than standard deviations)
Hope this helps
Alfonso
Originally posted by Diana Parvinchi:
Your previous post seems to contain only one attachment (named DifferencesContrasts_ASD1_ASD2_ASD3_Social_Interpersonal.png and containing a figure labeled "connectivity between SPL l and tolTG l"). The NITRC forum only allows you to attach a single file to a post, could you please either zip your multiple screenshots into a single compressed file and attach that one, or send me by mail the original screenshot files?
In the meantime, regarding your other questions:
1) yes, you may simply select all of the ROIs typically associated with that network and either enter a between-sources contrast "eye(N)" (where N is the number of ROIs; e.g. [1 0 0;0 1 0;0 0 1] for three ROIs), which will test your effect across any of the regions in the network instead of on a single region, or more typically enter a between-sources contrast "ones(1,N)/N" (e.g. [1/3 1/3 1/3] for three ROIs) which will test your effect when looking at the average connectivity with all of the regions in the network instead of a single region. note: the between-subject effects / contrasts would be just the same as before, you would only be modifying the selected sources and the associated between-sources contrast
2) I am not entirely clear if this is what you mean here, but you can right-click on any results table (or ctrl-click if on a Mac) and export that as a text file, as a .csv spreadsheet, or as a .mat Matlab file. If, instead, you mean modifying the barplot displays, you could type the following in Matlab command-line "set(gcf,'menubar','figure')" (without the double-quotes) and then use the "Edit" options there to modify these displays, or you could use the "import values" button on the main CONN gui to get the individual connectivity values for each subject if you prefer to analyze those separately and create your own displays, for example.
3) The error bars are 90% confidence intervals for the parameter estimates (those are typically larger than standard errors and smaller than standard deviations)
Hope this helps
Alfonso
Originally posted by Diana Parvinchi:
Hi Alfonzo,
Thanks again for your help. I am wondering if you've had a chance to take a look at our attachments. I re-sent you three screen shots of different tables to get your input on each. In all the tables, the first three regressors are the three age-groups (age-group1: 3-7 years; age-group2: 8-12; age-group3: 13-20 years) and the last three regressors are scores on a clinical measure (social interpersonal relationships) for each group. I used the between-subjects contrast you specified previously [0 0 0 1 -1 0; 0 0 0 0 1 -1] for two of the tables and I used the any-effect contrast from the options listed under the Between-subjects contrast for one table. Could you please provide your interpretation of these tables? The table with the any-effect is labels as such and I'm very curious to get your input on that - not sure what this contrast means in this context.
I also have a couple of other questions:
1) If we're interested in connectivity within a network (e.g. default mode), could I select several ROIs corresponding to the regions making up that network and run the same contrasts?
2) could we make changes to the tables?
3) Are the error bars in the tables Standard Deviations or Standard Errors - all very large in our case?
Many many thanks for your help.
Best,
Diana.
Originally posted by Diana Parvinchi:
Thanks again for your help. I am wondering if you've had a chance to take a look at our attachments. I re-sent you three screen shots of different tables to get your input on each. In all the tables, the first three regressors are the three age-groups (age-group1: 3-7 years; age-group2: 8-12; age-group3: 13-20 years) and the last three regressors are scores on a clinical measure (social interpersonal relationships) for each group. I used the between-subjects contrast you specified previously [0 0 0 1 -1 0; 0 0 0 0 1 -1] for two of the tables and I used the any-effect contrast from the options listed under the Between-subjects contrast for one table. Could you please provide your interpretation of these tables? The table with the any-effect is labels as such and I'm very curious to get your input on that - not sure what this contrast means in this context.
I also have a couple of other questions:
1) If we're interested in connectivity within a network (e.g. default mode), could I select several ROIs corresponding to the regions making up that network and run the same contrasts?
2) could we make changes to the tables?
3) Are the error bars in the tables Standard Deviations or Standard Errors - all very large in our case?
Many many thanks for your help.
Best,
Diana.
Originally posted by Diana Parvinchi:
Hi Alfonzo,
I'm sorry about that - not sure what happened. I have logged into the system and attaching the files from the website instead of e-mail. I hope this method will work. Thanks again for all your help!!!
I'm sorry about that - not sure what happened. I have logged into the system and attaching the files from the website instead of e-mail. I hope this method will work. Thanks again for all your help!!!