[pymvpa] Group level RDMs (Nick Oosterhof)

[Nick Oosterhof]

> On 25 Feb 2016, at 16:32, Martin Sjøgård <sjogard at stud.ntnu.no> wrote:
> 
> Ok thanks for the suggestion. But does distatis average all subjects' response patterns before then making an RDM for the averaged data, or make one RDM per subject and then average across RDMs? This is crucial, since the former could destroy individual idiosyncratic patterns.

With DISTATIS, the RDMS for each participant is used as the input. The DISTATIS method involves an 'optiomal' weighting of each RDM, so that participants that agree most with the group average are given a larger weight.