Hi Mark,
I'm not quite sure if I understand you correctly: if in the feat gui
you select 'first-level analysis', the data selection tab should give
you a button 'Select 4D data'. Are you saying that then this selection
accepts a feat directory as input and then picks the filtered_func_data
automatically? I just tried on my mac and can't reproduce this. The
normal behaviour is that if there is a feat directory 'blah.feat' which
is specified as the 'Select 4D data' input, feat can do one of two
things: (i) though it does not complain when pressing 'OK', the 'total
volumes' counter will remain at 0 (because it didn't find any data)
(ii) if there is 4D data 'blah.hdr' and 'blah.img' available in the
same location where the 'blah.feat' folder is located, feat will use
this as input. In this case, however, the results should not look
different (provided the feat folder 'blah.feat' was created by running
feat on 'blah.[img,hdr]').
Alternatively, did you use the filtered_func_data within a feat
directory for a new lower-level-analysis? If so, you will need to
switch off the pre-processing steps, otherwise the data will be
smoothed again (etc) and results will almost definitely look very
different....
It's probably best to have a closer look at the report.com and
report.log files within the different feat folders to find out what's
been going on. If you want to email these directly, I'm happy to check.
all the best
christian
On 21 Apr 2004, at 23:20, Mark Cohen wrote:
> In a recent data analysis, we re-ran feat processing at the single
> subject level (in order to change
> the registration target only) and were surprized to find that the
> resultant statistical data were
> much altered. The re-analysis was set up using the FEAT directory as
> input. Our expectation was
> that this would simply re-run the GLM and registration based on
> filtered_func data. Instead, from
> the report.com file, it appears that feat also repeated the smoothing
> on filtered_func, which had
> dramatic effects on the stats.
>
> Is this the expected behavior or a bug? Is it permissible to repeat a
> first level analysis using the
> directory as input, or should we manually select the input files.
>
> Thanks.
>
--
Christian F. Beckmann
Oxford University Centre for Functional
Magnetic Resonance Imaging of the Brain,
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
Email: [log in to unmask] - http://www.fmrib.ox.ac.uk/~beckmann/
Phone: +44(0)1865 222782 Fax: +44(0)1865 222717
|