 |
 |
2009/5/11 Daniel Krieger <[log in to unmask]> > Hi; > > Many thanks. Just to follow up, would the following be valid: > > A-(BandC): 1 -1 -1 > B-(AandC): -1 1 -1 > C-(AandB): -1 -1 1 > AandC (to determine where both show activation): 1 0 1 > AandBandC (to determine where all show activation): 1 1 1 > > or would I need to use contrast masking to achieve this? Sorry if this is > elementary stuff. > > These contrasts may not show exactly what you want. e.g. 1 -1 -1 will show regions where the average of B & C is significantly less than A. 1 0 1 will show regions where responses corresponding to A & C are significantly greater than 0. If you want to show regions where two or three different contrasts are jointly significant, use contrast masking. > Daniel > > On Mon, May 11, 2009 at 3:08 PM, Eugene Duff <[log in to unmask]>wrote: > >> Hi Daniel, >> The second model is appropriate here. The tripled t-test is necessary if >> you want to include regressors modelling another factor like the mean >> response levels of individual subjects, which will account for the group >> mean response level. In your first model you are only modelling the >> differences between conditions, with nothing modelling the overall mean. >> Cheers, >> Eugene >> >> 2009/5/11 Daniel Krieger <[log in to unmask]> >> >>> Hi All, >>> >>> (Newbie, so I apologize in advance) >>> >>> So here is my scenario: >>> Single subject, 9 total runs (3 with condition A, 3 with condition B, 3 >>> with >>> condition C) all collected during a single visit. >>> >>> 1st Level Analysis on each run separately. >>> - 1 EV: Simple 20s block design (rArArArA or rBrBrBrB or rCrCrCrC) >>> - 1 Contrast: 1 >>> >>> 2nd Level Analysis (here is the quandry) >>> Should I use? >>> Setup 1 (similar to "tripled" t-test" on website but without additional >>> EV >>> for removal of means) >>> EVs: >>> Input Grp EV1 EV2 >>> A1 1 1 1 >>> A2 1 1 1 >>> A3 1 1 1 >>> B1 1 -1 0 >>> B2 1 -1 0 >>> B3 1 -1 0 >>> C1 1 0 -1 >>> C2 1 0 -1 >>> C3 1 0 -1 >>> >>> Contrasts EV1 EV2 >>> A-C 1 2 >>> C-A -1 -2 >>> A-B 2 1 >>> C-B 1 -1 >>> B-A -2 -1 >>> B-C -1 1 >>> A mean 1 1 >>> B mean -1 0 >>> C mean 0 -1 >>> >>> Or should I use setup 2? >>> EVs: >>> Input Grp EV1 EV2 EV3 >>> A1 1 1 0 0 >>> A2 1 1 0 0 >>> A3 1 1 0 0 >>> B1 1 0 1 0 >>> B2 1 0 1 0 >>> B3 1 0 1 0 >>> C1 1 0 0 1 >>> C2 1 0 0 1 >>> C3 1 0 0 1 >>> >>> Contrasts EV1 EV2 EV3 >>> A-C 1 0 -1 >>> C-A -1 0 1 >>> A-B 1 -1 0 >>> C-B 0 -1 1 >>> B-A -1 1 0 >>> B-C 0 1 -1 >>> A mean 1 0 0 >>> B mean 0 1 0 >>> C mean 0 0 1 >>> >>> Or am I missing the boat entirely? I've tried running higher level >>> analysis >>> with both set-ups and the results come out nearly but not exactly >>> identical >>> except for the contrasts to determine condition means. Alternatively, >>> should >>> I run a higher level for each condition separately (single group average) >>> and then feed those into a third level using one of the above setups but >>> with only 3 inputs? >>> >>> Many thanks, >>> >>> Daniel >>> >> >> >> >> -- >> >> Eugene Duff >> >> FMRIB Centre, >> University of Oxford >> John Radcliffe Hospital, Headington OX3 9DU Oxford UK >> >> Ph: +44 (0) 1865 222 739 Fax: +44 (0) 1865 222 717 >> >> -- >> > > -- Eugene Duff FMRIB Centre, University of Oxford John Radcliffe Hospital, Headington OX3 9DU Oxford UK Ph: +44 (0) 1865 222 739 Fax: +44 (0) 1865 222 717 --