F value is the ratio of the sum of squred residuals between the null
and alternate models so it can't possibly be negative. Recheck your
code and computation.
On 5/21/05, Daniel H. Mathalon <[log in to unmask]> wrote:
> Dear SPMers,
>
> I have attempted to implement a second level ANOVA (within subjects)
> with 19 subjects and 5 conditions. I chose to do a sphericity
> correction and identified the conditions as correlated measures.
>
> The resulting SPM_F image for "all effects of interests" contain
> negative values at all voxels, with zero being the highest value.
> Negative F values should not be possible, should they?
>
> I also specified a t-contrast with weights for the 5 conditions of -2
> -1 0 1 2. The resulting t-map contains uniform values of zeros.
>
> Any ideas about what may have gone wrong and how I can trouble-shoot
> these peculiar results would be most appreciated.
>
> Thanks.
>
> Dan
>
--
Moo Chung
http://www.stat.wisc.edu/~mchung
Assistant Professor
Department of Statistics
University of Wisconsin-Madison
|