Hi Nathan,
>I noticed that the surf plots of the parametric responses were the same
>across sessions. Exploring this further in spm_graph.m around line 495:
>
>B = beta(Sess(s).Fc(u).i);
>
>The variable B is assigned the same two beta values (for the main effect and
>parametric columns of xX.X, respectively) regardless of the value of
>variable s (i.e. session). Has anyone seen this before? Or is this more
>likely due to my ignorance in setting up the SPM.Sess.U.P.i fields
>for multiple session studies....
It does seem that you've found a bug. In fact, the plots will not always
look the same, even if the betas used to plot them are identical (if the
parametric modulators are different in the two sessions, for example), so I
can see how this slipped throug the net!
Changing this line to:
B = beta(Sess(s).col(Sess(s).Fc(u).i));
should correct this error. I hear from Karl that this bug will also affect
Volterra plots, so anyone who uses them might be advised to change line 546
similarly, and also line 581, I think. I hope that a bug fix will soon be
posted.
>Also, is it ok to plot the "average" parametric response across sessions?
>This doesn't seem to be a current option in SPM2, but I could envision just
>averaging the fitted responses Y over sessions and invoking the surf plot,
>surf(pst,P,Y'). Yes/no/maybe?
I wouldn't simply average the plots if I could help it, but the "correct"
approach might depend to some degree on the nature of the parametric
modulator. I think that a better approach would be to calculate the
average parametric modulator (across sessions) for level of the event and
the average parameter estimate for each session. You can then plot an
averaged fitted model using the code in spm_graph.m - you will have to have
some matlab familiarity to use this solution. For some experimental
designs the straightforward average and my suggested approach might
converge. (My approach can also potentially be used to plot average fitted
responses across subjects.)
>Thanks for all the help in advance!!
Hopefully this falls into the category of help...
Well bug-spotted!
Best wishes,
Joel
_______________________________________________
Joel Winston
PhD Student
The Wellcome Department of Imaging Neuroscience
12 Queen Square
London WC1N 3BG
Tel: +44 (0) 20 7833 7477 <-- Note new number!
Fax: +44 (0) 20 7813 1420
Web: http://www.fil.ion.ucl.ac.uk/~jwinston
_______________________________________________
|