 |
 |
above 2012/7/5 Auer, Tibor <[log in to unmask]> > According to the randomized manual ( > http://www.fmrib.ox.ac.uk/fsl/randomise/index.html), -D option is > necessary if you are not modelling the mean in the design matrix. Since you > first EV is the group mean, you do not need -D option. After reading it, I > am not sure if it also demeans the design or only the data.** > > ** ** > > I am not sure what do you mean “leave the –D option”. Leave it in the > command, or leave it out? Sorry, I mean leave in the command I did't > remove it and I tried 1 and 2.** > > ** ** > > If you want to be on the safe side, enter demeaned values into EV2, and > leave the -D option out of the command? so you mean a design matrix with > FA EV1 (not demeaned data) and EV2 (demeaned data)? but I would not remove > -D if there is the chance that it serve to demean the design.** > > ** ** > > One possible explanation for your “increased” significance in your case 1, > is that, it also contains your group mean, since both your EVs incorporate > group mean. As a test, I would contrast for EV1 [1 0]; and in your case 1 > (demeaned EV2) I would expect lower significance for this contrast.**** > > ** ** > > *Auer, Tibor M.D. Ph.D*.**** > > Biomedizinische NMR Forschungs GmbH**** > > am Max-Planck Institut für Biophysikalische Chemie**** > > Am Fassberg 11**** > > 37077 Göttingen**** > > Germany**** > > Phone/Work: +49-(0)551-201-1725**** > > Phone/Home: +49-(0)551-387-0076**** > > Mobile: +49-(0)176-8012-7921**** > > Mail: [log in to unmask]**** > > ** ** > > *From:* FSL - FMRIB's Software Library [mailto:[log in to unmask]] *On > Behalf Of *mauro serra > *Sent:* Thursday, July 05, 2012 11:47 AM > *To:* [log in to unmask] > *Subject:* Re: [FSL] demean in TBSS**** > > ** ** > > Hi all,**** > > I tried to leave the -D option and either:**** > > 1) in one case insert row data in EV**** > > 2) in the other case insert demeaned data in EV**** > > the result were really similar but in the case of 2 there was a bit less > significance.**** > > this seems to point in favor of Angela, what do you think?**** > > ** ** > > 2012/7/5 Angela Favaro <[log in to unmask]>**** > > Hi ALL, > I think you have to demean both: > 1) behavioral data (EV1) > 2) image data (using -D) > > Why does it make no sense? Demeaning EV1 and -D are different processes > (one for behavioral data and the other for image data) > > Angela**** > > > > > Hi, > > > > Demeaning (manual or automatic) is necessary, since you model the mean in > > EV1 > > already. So you have only two options: > > 1. either you enter raw data and use the -D option (automatic) > > 2. or you enter demeaned data and you do not have to use -D option > > (manual) > > (It makes no sense - however, make no harm either - using -D option on > > demeaned data, since the mean is 0. As a result, you demeaned data will > > be > > untouched.) > > > > I hope it helps, > > > > > > Auer, Tibor M.D. Ph.D. > > Biomedizinische NMR Forschungs GmbH > > am Max-Planck Institut für Biophysikalische Chemie > > Am Fassberg 11 > > 37077 Göttingen > > Germany > > Phone/Work: +49-(0)551-201-1725 > > Phone/Home: +49-(0)551-387-0076 > > Mobile: +49-(0)176-8012-7921 > > Mail: [log in to unmask] > > > > > >> -----Original Message----- > >> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On > >> Behalf Of Mauro Serra > >> Sent: Wednesday, July 04, 2012 5:36 PM > >> To: [log in to unmask] > >> Subject: [FSL] demean in TBSS > >> > >> HI, > >> > >> I'm new in fsl DTI analisys, I'm not shure if I have understood when I > >> need > > to > >> demean my data. > >> > >> I'm performing a TBSS analysis within subjects, with only one EV: I'm > >> correlating FA and a behavioural measure. > >> > >> I have posted the design matrix and the contrast matrix, my questions > >> are: > >> > >> 1) Should I use randomize with the option -D like in this script: > >> randomise -i all_FA_skeletonised -o tbss -m mean_FA_skeleton_mask -d > >> design_nco.mat -t design_nico.con -n 500 -D --T2 or not? > >> > >> 2) if I have to use -D should I insert in the design matrix the > >> behavioural > >> demeaned data? or should I insert the "row" data and the -D option will > >> do > >> everything automatically? > >> > >> 3) if I don't have to use the -D option in the script, what about the > >> behavioural data? should I write in the design matrix the demeaned > >> behavioural score or row score? > >> > >> Thanks in advance for the help, and I'msorry if my question looks > >> stupid > > but > >> I'm really stuck on it! > > > >**** > > > > **** > > ** ** > > -- > Mauro.**** > -- Mauro.