Dear Fahimeh,
Did you use the same table of model evidence values for each of the three runs of spm_BMS_gibbs.m (versions 4010, 4290 and 4667) ?
If so, can you send me the table and I'll look into it.
Best,
Will.
> -----Original Message-----
> From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
> On Behalf Of Fahimeh Mamashli
> Sent: 06 June 2013 20:26
> To: [log in to unmask]
> Subject: [SPM] BMS different results
>
> Dear SPM experts,
> last year, I have computed DCM for MEG and I have done two types of
> family inference Bayesian model selection: first, based on connectivity
> types, second: based on the number of regions.
>
> now I am repeating the same analysis using different versions of SPM
> and surprisingly I found totally different results with each version!
>
> e.g. with SPM 4010, I found similar result as last year for the family
> inference based on the connectivity types but totally different result
> for the second family inference(based on the number of regions).
>
> with SPM4010 but only updating the spm_BMS_gibbs.m (attached to the
> email), I found similar result for the second family inference but
> totally different first family inference!!
>
> with SPM4290, I found similar result as SPM4010.
>
> with SPM4667, the result is similar to SPM4010 and updated
> spm_BMS_gibbs.m
>
> I am very confused and don't know which result is really true?? does
> anyone knows how I can find this out?
>
> As far as I have learned from the spm emails, this problem is due to a
> bug in spm_BMS_gibbs.m but the question is when the result is due to a
> bug and when not.
>
> I am going to do BMS for a new dataset but I am very unsure about the
> BMS result as it turns out to be unstable.
>
> Thank you very much for your time.
>
> I am looking forward to hear from you.
>
> Best wishes,
> Fahimeh
> ------------------------
> PhD student
> Max Planck Institute for Human Cognitive and Brain Sciences
> Stephanstraße 1A
> 04103 Leipzig
> Germany
>
> Tel: +49 341 9940 - 2570
|