 |
 |
Dear Francis, > 1. Uncorrected p value. > > I am having some difficulty determiniing what is acceptable in terms of > an uncorrected p value. For example lets say I set the voxel threshold to > an uncorrected level of p<0.01 and get a list of regions, most of which I > would predict. Typically The areas I am interested in are between the > range of p<0.009 and p <0.0001 (uncorrected). When I set this to > p<0.001 I obviously loose some of the areas. Is using a threshold of p > <0.01 (uncorrected) level acceptable or should I use p<0.001 I would report the results at two levels. First go through the anticipated regions using a SVC (at p <0.05). Second report anything you had not predicted (but is interesting) 'descriptively' at p <0.05 uncorrected. The second reports can be used by you or others to guide anatomically specific hypotheses in future experiments. Say at the beginning of the results that you will be reporting the results at these two levels. > My next step is to focus on the areas of interest, basal ganglia (e.g > globus pallidus) or frontal cortex and use the SVC option. When I run > the SVC do I use the voxel or cluster level (corrected) p value ? Use a voxel-level SVC p value. > 2. SVC > > In one of your previous emails to me you suggested using a SVC with a > radius of 16 mm for anticipated areas and a correction for the whole > volume for non anticipated areas. > > In my case the anticipated areas are cortical (eg parietal or frontal > gyri) and sub-cortical (basal ganglia) and as such will have different > shapes and volumes. Is it appropriate therefore to use a fixed value > (16 mm) for all these regions or should I use different radii for each > of the different regions ? I have been using the sphere option and > specifying a 16 mm radius for the sub-cortical regions but am unclear > what to use for the cortical regions. In one of your previous papers, > J. Neuroscience Dec 1999 p 1087, you based your corrections on the > volume of interest and cited Filipek et al 1994. and Worsley et al > 1996. I have the Worsley paper and know there is a table that lists the > cc values of different areas. I think you have the lattidude to change the radius depending on the structure involved. You could use any published data as a guide. The Caudate, for example, would have a small radiou (sqy 6mm) whereas parietal cortex should be greater unless you specify which part (e.g. IPS). I hope this helps - Karl %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%