you should set "Variance" to "Equal" in your two-sample T-test. The
difference is due to a change in the ReML algorithm as described in an
earlier thread (that I don't manage to find right now).
Ylva Lilja wrote:
> I recently down loaded SPM8, with the latest updates, and tried to run a
> two-sample t-test. Everything works fine up until the "Result" part when
> I try to construct the contrasts. I write them exactly the same way as I
> have been doing it in SPM8b, for example 1 -1 0 0 for a t-test with two
> groups and two covariates. I get an error message when I do this
> however, saying invalid contrast. I also notice that the design matrix
> in the SPM contrast manager looks funny: in the first column, with group
> 1 images, all the cells are gray, as if there are no images entered for
> that group. The design matrix that appear before Estimation looks good,
> Does anyone know if something changed in this step compared to
> SPM8b/SPM5 or if there is perhaps a bug?
> Very grateful for help!
> Ylva Lilja
> UC Irvine
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG