Hello,
I am interested in computing the log evidence (F) of the Bayesian deconvolution of the VOI time series for an PPI approach with simple deconvolution.
This is a simple hierarchical model with two levels, as described in Gitelman et al., 2003.
In order to compute F, I tried to use the SPM functions: spm_peb_ppi.m and spm_PEB.m.
To the best of my knowledge, the log evidence is only computed providing spm_PEB is called with two or more output variables as follows (just copied here for simplicity),
if nargout > 2
% condotional covariance of h
%----------------------------------------------------------------------
Ph = -dFdhh;
% log evidence = F
%----------------------------------------------------------------------
F = - Py'*Ce*Py/2 ...
- length(I{1})*log(2*pi)/2 ...
- spm_logdet(Ce)/2 ...
- spm_logdet(Ph)/2 ...
+ spm_logdet(hP)/2 ...
+ spm_logdet(Cby)/2;
end
Unfortunately, the value of F is not computed because the variables dFdhh (and Ph) and Py are not defined.
Does anyone know the reason for that? Is it possible to get the log evidence for the deconvolution?
Thanks very much in advance,
Cesar
Cesar Caballero Gaudes
CIBM / Radiologie
Rue Gabrielle-Perret-Gentil 4
(ex-rue Micheli-du-Crest 24)
CH-1211 Genève 14
Switzerland
[log in to unmask]Phone: (+41) 0223725217
