 |
 |
Hi Anderson, I have the same issue as Laura: there are 2 networks that are part of the default mode network that have come out of the ICA. Following this, I ran dual_regression. Before I saw this email thread I was thinking of two ways to analyze the data (below), but it seems the approach you are scripting may be the best approach. Can you please help me understand why these approaches are not sound? Sorry in advance for my very basic questions! 1) Can one simply add the two networks using fslmaths. Adding two components of the dual regression like this seems to work: fslmaths dr_stage2_ic0006.nii.gz -add dr_stage2_ic0002.nii.gz result_file Following this, I would plan to carry on with randomise. Perhaps there are reasons why this approach is not valid that I do not yet understand? 2) Can one do statistical analyze between groups of the two components independently without ever combining them? If I do this, can't I just correct for multiple comparisons in the stats phase and get valid results? Thanks a million! Sheena On Tue, Jun 7, 2016 at 8:56 PM, Anderson M. Winkler <[log in to unmask]> wrote: > Hi Laura, > > Please give me a few days to prepare a script for this. I'll try to send > details soon. > > All the best, > > Anderson > > > On 6 June 2016 at 19:51, Laura Korthauer <[log in to unmask]> wrote: > >> Hello, >> >> I ran group ICA and got two default mode network components, one that is >> more anterior and one more posterior. I'd like to combine them into one >> single component. From a previous listserv response that Anderson made to >> another user with a similar problem, I see that: >> >> "Another option to consider would to multiply these two components by >> their respective lines in the mixing matrix (that is, mix them back, >> "undoing" the ICA for them), then run ICA again only on this "undone" data, >> with dimensionality 1." >> >> I'm confused about how exactly to go about this. The mixing matrix has 20 >> columns (corresponding to my 20 components, presumably) and 30444 rows (129 >> subjects x 236 vols). What files am I to be multiplying to mix them back >> and "undo" the ICA? >> >> Thanks for your help, >> >> Laura >> >> -- >> Laura Korthauer >> Ph.D. Candidate, Clinical Psychology >> University of Wisconsin-Milwaukee >> > >