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
---------------------------------------------------------------------------
|