 |
 |
Dear Daniel, as mentioned by Raphael, the change in the order of the columns of the design matrix is expected and has, by itself, no impact in the inference, provided that your re-order your contrasts accordingly. However you might also get a difference in non-sphericity estimation (because a different set of voxels will be selected) and consequently different results - I would consider r4290 to be more accurate. On the design matrices you sent, the effect of the whitening seems to be similar and I wouldn't expect large differences between the two versions. It is possible to use SPM.mat files specified/estimated from r4010 in r4290. Best regards, Guillaume. On 20/04/11 12:14, Raphael Hilgenstock wrote: > Hello Daniel, > > no reason to worry, just take a look a the release note accompanying the > latest update of SPM8 (4290): > > http://www.fil.ion.ucl.ac.uk/spm/software/spm8/SPM8_Release_Notes_r4290.pdf > > In fact, only the subject factor has been shifted/moved to the > right-hand side of the design matrix (apart from fixing a bug). > > Moreover, once the design is estimated (e.g.) using release 4010 there > is no need to reestimate the model for displaying results of the very > same model/design using release 4290 (and vice versa), at least on my > machine. Just select the model (using release 4290) that is already > estimated (by release 4010) and choose the contrast of interest. Could > the differing results between both releases be due to the fact that you > did not adapt the contrasts to the new order of factors (subject, > condition, group) in the design matrix of the new release (take a close > look at the design matrix)? > > Hope I could help and best regards > Raphael > > > Am 20.04.2011 12:07, schrieb Daniel Ferreira: >> Dear SPM experts, >> >> I would very appreciate if you can help me. >> >> Is something different in the spm8_4290 version regarding the design matrix/estimation as comparing with the spm8_4010 version? >> >> The thing is that I have tried to estimate the same identical flexible_factorial.mat in both versions (spm_4010 and spm_4290). I do get different design matrices (mirrored ones) and I do not get why and of course the results are completeley different. >> >> Can I avoid to create a new .mat for the 4290 version? I mean, can I use the 4010's one in 4290 version? >> >> Please find attached the design matrices. One has 6 conditions, the other one has 66. >> >> >> Thank you very much in advance, >> >> Daniel Ferreira >> > > -- Guillaume Flandin, PhD Wellcome Trust Centre for Neuroimaging University College London 12 Queen Square London WC1N 3BG