Dear Tobias,
Good to hear that you are enjoying your new life at Princeton! I hope you have had a good start.
There are two ways in which you could proceed without having to re-estimate your GLMs:
1. For each subject, you could combine the parameter estimates across sessions using the "average" function in DCM. For each parameter estimate, this will give you a weighted mean across sessions (weighted by the session-specific posterior variances). You could then feed the resulting values into a classical second level analysis (e.g. t-test).
2. You can account for the dependency amongst the parameter estimates within each subject by using a repeated measures ANOVA ("repeated measures" because you have multiple parameter estimates for each subject, i.e. one for each session).
All the very best,
Klaas
----- Ursprüngliche Mail ----
Von: Tobias Overath <[log in to unmask]>
An: Klaas Stephan <[log in to unmask]>
Gesendet: Dienstag, den 12. Februar 2008, 17:18:57 Uhr
Betreff: DCM with multiple sessions
Hi Klaas,
I hope you are enjoying your life together in Zurich - I am very
happy here in Princeton, sharing life with Lillian at last.
I have a hopefully quick question regarding multiple sessions in DCM.
I understand one way is to concatenate the session (and introduce
session constants etc.); however, I would rather avoid that (mainly
for time reasons, i.e. having to re-estimate the GLMs). Instead, I
have estimated each model for each session separately (e.g.
DCM_model1_sess1, DCM_model1_sess2, DCM_model1_sess3,
DCM_model1_sess4). I have searched in the SPM archives and in your
Powerpoint slides to find a way how to combine the different session
DCMs of the same model?
Or would you suggest I bite the bullet and re-estimate my GLMs with
concatenated sessions?
Best,
Tobias
Heute schon einen Blick in die Zukunft von E-Mails wagen? Versuchen Sie´s mit dem neuen Yahoo! Mail. www.yahoo.de/mail
