 |
 |
Hi - sounds like you have bad outliers - I would start by looking at the "filtered_func_data" and "var_filtered_func_data" inputs to the higher-level analyses, that are stored in the ---.gfeat/cope*.feat directories. Load these into FSLView, turn on timeseries viewing, and check for clear outlier subjects. Cheers. On 3 Jun 2009, at 08:42, Raymond Salvador wrote: > Dear Steve, > > Thanks for your quick answer. There is no contrast called "error". I > have checked the log files for each high level cope and, in the > iterations it is full of "ndtri domain error" followed by message: > > WARNING: The passed in varcope file, tmpvarcope00.., contains voxels > inside the mask with zero (or negative) values. These voxels will be > excluded from the analysis. > > The output images are with extensive yellow "activations". > > Thanks again > > Raymond Salvador > > > > 2009/6/3 Steve Smith <[log in to unmask]> > Hi - If this message appeared at the top of your HTML report, it is > _possible_ that this has just been picked up because it has found > the word "error" in one of your contrast names - is that possible? > If not - then you need to look through the more detailed log page to > see where the problem has occurred. > > Cheers. > > > > On 2 Jun 2009, at 11:40, Raymond Salvador wrote: > > Dear FSL staff, > > I'm running a full factorial 3X2 ANOVA and I keep on having the same > message in the output: > > "Errors occurred during the analysis" > > Although the model is unbalanced, it contains several individuals in > each of the 3x2 = 6 levels. I think the dessign matrix is OK. I > started from a full matrix with redundant columns > > mean P1 P2 G1 G2 G3 P1_G1 P1_G2 P1_G3 > P2_G1 P2_G2 > P2_G3 > 1 > 1 0 1 0 0 1 0 0 0 0 > 0 > 1 > 1 0 0 1 0 0 1 0 0 0 > 0 > 1 > 1 0 0 0 1 0 0 1 0 0 > 0 > 1 > 0 1 1 0 0 0 0 0 1 0 > 0 > 1 > 0 1 0 1 0 0 0 0 0 1 > 0 > 1 > 0 1 0 0 1 0 0 0 0 0 1 > > Where P and G are factors with 2 and 3 levels (for simplicity I only > show one row per combination of levels, but there are as many rows > as individuals). > > Then, I have eliminated columns to obtain a full rang matrix (with 6 > linearly independent columns) > > mean P1 G1 G2 P1_G1 P2_G3 > 1 1 1 0 1 0 > 1 1 0 1 0 0 > 1 1 0 0 0 0 > 1 0 1 0 0 0 > 1 0 0 1 0 0 > 1 0 0 0 0 1 > > This is the design matrix I've used (but moving the mean column of > ones to the end). I have set a contrast for each column (EV) but the > last one (coding for the mean). > > C1 1 0 0 0 0 0 > C2 0 1 0 0 0 0 > C3 0 0 1 0 0 0 > C4 0 0 0 1 0 0 > C5 0 0 0 0 1 0 > > And I've asked for three different F tests > > 1) One for C1 (main effect of factor P [with two levels]) > 2) One for C2 and C3 (main effect of factor G [with three levels]) > 3) One for C4 and C5 (columns coding for interaction) > > I haven't centered any column as (I understand it) centered and > uncentered vectors span exactly the same linear subspaces, and they > are equivalent in an F-test comparing the full model with a nested > submodel. Here I'm assuming that each required F-test is related to > a comparison between the full model and the submodel without the > columns of the contrasts included in the F-test (as it is usually > the case in standard GLM theory). Is that right? > > I've have used FLAME1 and Mixed OLS, with common variance estimate > for all groups (Group 1 for all), leading to the same error message. > > Sorry for this loooooonnngg question/explanation, and thanks very > much in advance. > > > Raymond Salvador > > > > > > --------------------------------------------------------------------------- > Stephen M. Smith, Professor of Biomedical Engineering > Associate Director, Oxford University FMRIB Centre > > FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK > +44 (0) 1865 222726 (fax 222717) > [log in to unmask] http://www.fmrib.ox.ac.uk/~steve > --------------------------------------------------------------------------- > --------------------------------------------------------------------------- Stephen M. Smith, Professor of Biomedical Engineering Associate Director, Oxford University FMRIB Centre FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK +44 (0) 1865 222726 (fax 222717) [log in to unmask] http://www.fmrib.ox.ac.uk/~steve ---------------------------------------------------------------------------