I've thought about this issue in the past, but not enough to have a definitive answer.
Here's a post I made previously which cited former SPM stats guru Andrew Holmes:
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=SPM;fadc02b2.1109
In particular, AH wrote, "The statistic (SPM) images (SPMt_????.img & SPMF_????.img) should *not* be entered into a second level analysis if you want to effect a random effects analysis. This would basically be assessing the significance (across subjects) of the individual subjects significance! (Rather than the significance (across subjects) of the response."
I think the basic reasoning is that we're interested in drawing inferences on the effect size, not on the significance.
As others have stated, I'm pretty sure that if you use Z scores rather than the contrast itself, you're not going to be doing random effects.
Perhaps the way to settle the issue would be to write down a bunch of models and compare what this method estimates to what you _want_ to estimate. (That would show, for example, that it's not measuring the between-subject variation.)
If you're really dying to know you could try asking on one of the sci.stat.* newsgroups. :-)
|