Hopefully someone can provide me with some much needed advice.
I was wondering about the possibility of using 2nd order polynomials in the general linear model in SPM8.
I (think I) would like to do a VBM analysis with this method, if it is possible?
If I take, for example, ratings of the frequency of delusions as the variable of interest using a response scale of 1-10. How could I implement a model to test if rather than a linear relationship between grey matter in various brain regions and delusions, there is a relationship were e.g. up until a point there is not very much decrease in grey matter at all, but beyond this there is a steeper decrease in grey matter e.g. as the ratings increase through 7, 8, 9, 10.
If I include the actual ratings in the 1st column, and square the values and add these to the 2nd column (which I think is the method Buchel et al (1996) describe) - is this appropriate? Or is it much more complicated than this?
I would like to include the usual VBM covariates.
I would also like to look for decreases and increases in GM with this method (steeper increases low on the scale, steeper increases late on the scale, steeper decreases early on the scale, steeper decreases late on the scale). Would I have to reverse the responses on the scale e.g. replace the 1s with 10s, etc (and then square these) if I want to get all of these combinations?
Finally, what contrasts would I use to ask these questions?
If anyone has any advice it would be much appreciated.