> Many thanks for your immediate help.
>
> With your suggestions, I was able to run the stats
> successfully.
>
> After seeing the error message, initially I used
> threshold of 0 (I thought that was the lowest that can
> be used). I was not aware of using negative
> threshold. Does Inf means negative infinity?
Yes, so everything should be included. If you want to
increase sensitivity by reducing the search volume, then
I think you could use an explicit mask.
> I want to clarify a basic doubt: How to check for the
> image intensity in SPM?
You can use the Display button and click on various voxels.
This will then report the values in the lower left panel.
Note that the values are derived using the interpolation
method specified in the lower-right panel. Therefore, if
the images consist only of e.g. ones and zeros, there may
be other intensities shown because of interpolation.
> If I have understood correctly, the preprocessing
> steps will give difference image between early and
> late images.
Correct.
> After running the preprocessing steps in
> patients and controls can I use single subjects:
> conditions and covariates for the statistical
> analysis with age at baseline as a covariate using the
> modulated and smoothed difference images?
Yes. Note that you may also wish to include the time
interval between scans in the model, instead of the ones
and zeros that are usually used to denote different
conditions.
> For performing the stats is it a must that the number
> of difference images should be the same in patient
> (early and late) as well as control (early and late)
> groups?
This should not be necessary, although if your comparison
involves a single individual versus a group, then the stats
may be particularly susceptible to problems because of
any non-normally distributed errors. This effect can be
reduced by smoothing though.
Best regards,
-John
> What analysis threshold did you use? Check your
> images
> to see what intensity range you have, and select a
> threshold that is lower. If you used the "i1.*(i2-1)"
> expression, then most of the values will be around
> zero,
> so you should use a negative threshold (possibly -Inf,
> in order to include the whole FOV in the analysis).
>
> If your threshold is too high, then none of the voxels
> will be included in the analysis, giving you the error
> that you saw.
--
Dr John Ashburner.
Functional Imaging Lab., 12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491 or +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420 http://www.fil.ion.ucl.ac.uk/~john
|