Print

Print
I have made sure that I have the latest updates installed, and still I am getting the variance in the colour in the contant column as described below (and illustrated in my previous email).
 
the design matrix of some datasets exhibited a grey colour in the column for the constant while the majority were pure white as I would expect.  In other cases this column had soft grating (see attached image).  This was despite the fact that each particpant's design was specified using the exact same batch file.
 
Maybe I'm being silly and This is just reflecting larger variance in the mean voxel values over scans, but I have never seen this before nor have I seen it in example design matrices, so please let me know what it means and if it is normal.
 
thanks again

On Thu, Jun 4, 2009 at 1:01 PM, Gmail <[log in to unmask]> wrote:
This is a known old bug.  update your SPM installation.

-----
Darren Gitelman, MD
sent from my iPhone


On Jun 4, 2009, at 6:50 AM, Richard Binney <[log in to unmask]> wrote:

Hi,

I have previously run FMRI anlyses and not had any such problems, but during my latest analysis SPM5 is behaving strangely.

I'm currently specifying and estimating at the the first level (fixed single subject level).  When reviewing the design matrices follwioing specification, everything looked fine.  However, after estimation a few things have gone wrong:

1) Having previously estimated about 13 datasets, I specfied and estimated 4 newly acquired datasets.  2 strange things occurred: a) the SPM.mat file of a previously estimated dataset was overwritten so that in 'review' this SPM.mat file showed the design matrix (and associated images) of one of the newly estimated datasets; and b) the design matrix of some datasets exhibited a grey colour in the column for the constant while the majority were pure white as I would expect.  In other cases this column had soft grating (see attached image).  This was despite the fact that each particpant's design was specified using the exact same batch file.

2) In response to this, and to check it wasn't an error on my part (e.g. mixing file associations), I deleted all the first level analyses and completely re-specified and re-estimated all the datasets.  While the SPM.mat files seemed to have the correct associated files, the variance in the colour of the constant column (which was always pure white prior to estimation) got more severe than before.

3) I have also previously attempted to run two different group-level (random effects) analyses with this data  a onesample t-test and a two sample t-test.  This had a similar effect of mixing up associated SPM.mat and individual contrast files.

Can anyone tell me what is normal, what is not and what might be going on?

HELP!!

Thanks

Richard
<strangedesignmatrix.jpg>