Hi all, 

Sorry to keep posted queries regarding the meta-regression component of SDM. I do understand that those analyses should be treated as exploratory anyway. 

1. Does reporting associations based on peak voxels within a cluster inflate the overall effect size across the cluster? What are the grounds for using the peak, as per the GUI, rather than the ROI as a whole?
2. Relatedly, it seems that the primary test statistic being reported is based on whether the estimated intercept is significantly different from zero. Does the intercept represent the predicted z-value for a study coded zero for a particular IV? Could linear associations (i.e., betas) be reported instead? It does not seem that these are outputted. 

Many thanks in advance. 
Jack

Dear Jack,

the peak is the voxel of the cluster where effect size is higher and thus, as you say, it shows an effect size higher than the average effect size of the cluster. Note however that the effect size of the cluster cannot be straightforwardly estimated because voxels are not independent and thus their effect sizes cannot be simply averaged.

In a simple meta-regression, after scaling the independent variable to the [0,1] range, SDM software creates several z maps: the intercept, the slope, etcetera. Please see a description of them in the tutorial.

Hope this helps,