I realized coord anova doesn't do what you need, because it's more like a two sample t-test than a paired t-test (but you can have more than two groups).
It doesn't take the pairing into account. As far as I know, caret_stats doesn't do what you need it to do for the coords. If you can use something like local shape mancova (http://www.insight-journal.org/browse/publication/694) to generate your stat maps, then you can feed them to caret_stats' TFCE. But again, we're back to the randomization problem. I don't think that tool will generate the randomized stat maps using a strategy like those in Nichols & Holmes' primer paper. But you can do a paired diff on the depth maps. Just make sure the composites have the columns in corresponding sequence. Sorry to get your hopes up on the coords. On Apr 5, 2012, at 4:40 PM, Donna Dierker wrote: > I put three scripts here: > > http://brainmap.wustl.edu/pub/donna/US/UNC/ > login pub > password download > > For the coords, you'll need a different test -- coord anova: > > http://brainvis.wustl.edu/wiki/index.php/Caret:Documentation:Statistics#Coordinate_Difference_Analysis_of_Variance > > One of those scripts does that. But first, you'll need to x-flip either the > right or left. I propose x-flipping the left to right, resulting in mostly > positive values for x. See flip-x.sh. > > For your depth, you need to make a composite from your depth files, and the > gen_distortion.sh script does that for distortion. You'll want to recompute > your distortion using something like that, because the order of the mean > fiducial and the individual's fiducial is important. The mean is the first > argument to the distortion command. Else you end up with mostly negative > values, as you apparently did, and as I did when I first recomputed my term12 > version 2 distortion metric. Easy mistake to make, but we want to pump up > areas -- not shrink them down. > > Anyway, that script shows how to generate a composite and compute an average. > You only need the composite for your depth case, but the latter comes in > handy. > > (The gen_distortion.sh script didn't exist when it first came up the other > day. I just needed to do it, so now it does.) > > Donna > > > On Apr 5, 2012, at 10:17 AM, gangli wrote: > >> Hi Donna, >> >> I would like to test the paired difference between left and right >> hemispheres on 3D coordinate position and sulcal depth. Please find the >> uploaded input files and help me figure out how to do this. >> >> Thanks a lot. >> >> Gang >> >> >> My versions of caret_command can read them all, and the coords/metric all >> have roughly 164k nodes -- all good. >> But I still don't have your command line, or inputs like I'd expect to a >> paired t-test, if that's what you're doing. >> For each subject, I have a metric that has a column for 3D variability and a >> column for areal distortion. I doubt that you want to get a paired >> difference between these disparate measures. >> What you need, as input to the paired t-test, is a composite metric/shape >> with one column per subject. One composite might have left sulcal depth, the >> other right, for example. "It doesn't get any more paired than the left and >> right hemispheres of a person's brain." --Tom Nichols >> But it could be thickness at different timepoints, twins' thickness maps -- >> anything you can pair. But the two composite input files should have equal >> numbers of columns, and those columns should be paired (i.e., first col of >> each file corresponds to one another; second col of each file corresponds to >> one another; and so on). >> Hope this helps. >> On Apr 4, 2012, at 1:02 PM, gangli wrote: >>> Hi Donna, >>> >>> I have uploaded several generated subjects along with metric and shape >>> files. Can you please help me figure out how to average and compose the >>> shape and metric files? >>> >>> Thanks a lot. >>> >>> Gang >>> >>> Hi Gang, >>> Could you provide the full command line and/or output from running the >>> script? It might not work in Windows, but on Linux or MacOSX command line >>> you would do: >>> paired.sh >& paired.log >>> Then upload paired.log here: >>> http://pulvinar.wustl.edu/cgi-bin/upload.cgi >>> Depending on what information that gives me, I may want to see more >>> information, like your mean midthickness, distortion metric, composite >>> scalar files, etc. >>> Also, it would be helpful if when you reply you keep the prior message >>> history in your reply, so that I don't have to go to the archives to remind >>> myself of what you have already done. It is easy to get users confused -- >>> particularly when I am using this stuff heavily myself. >>> Thanks much, >>> Donna >>> >>> On Apr 4, 2012, at 9:07 AM, gangli wrote: >>>> I am using caret_distribution_Windows32.v5.65 and the input is *.metric >>>> file generated in previous discussion. Thanks. >>>> >>>> Regards, >>>> Gang >>>> _______________________________________________ >>>> caret-users mailing list >>>> caret-users at brainvis.wustl.edu >>>> http://brainvis.wustl.edu/mailman/listinfo/caret-users >>> _______________________________________________ >>> caret-users mailing list >>> caret-users at brainvis.wustl.edu >>> http://brainvis.wustl.edu/mailman/listinfo/caret-users >> >> _______________________________________________ >> caret-users mailing list >> [email protected] >> http://brainvis.wustl.edu/mailman/listinfo/caret-users > > > _______________________________________________ > caret-users mailing list > [email protected] > http://brainvis.wustl.edu/mailman/listinfo/caret-users _______________________________________________ caret-users mailing list [email protected] http://brainvis.wustl.edu/mailman/listinfo/caret-users
