Rebecca Gould wrote:
> Hi Will
>
>
>>BTW, I followed your advice and have been re-analyzing my RFX models
>>using bayesian inference. I've thresholded the data using a value of 0
>>as you suggested - now I've just got to decide what P value threshold I
>>should use for the PPMs in order that I can sensibly infer that there
>>
> really >are no activation differences between my 2 groups?
>
> "A good test is to see if area between plus and minus sigma is greater than
> 95% (where sigma is the 'background noise level' - otherwise known
> as the prior standard deviation)."
>
> Thanks for your response but I'm not quite sure how I would implement this
> in SPM. Any suggestions?
>
You could plot a PPM with activation threshold set as
before (eg. to the default prior s.d) and then
choose the 'p-value threshold for PPM' to be 0.05.
If your voxel of interest *does not* come up then with
probability 0.95 the voxel has not activated (ie. the
response - difference between groups - is less than threshold with probability 0.95).
Take care,
Will.
--
William D. Penny
Wellcome Department of Imaging Neuroscience
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7478
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
--
William D. Penny
Wellcome Department of Imaging Neuroscience
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7478
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|