Darren -
> As an addendum to Rik's question I'd like to add that the calculation
> of the se does not seem to take into account averaging across
> sessions (when plotting event related responses, fitted response +/-
> standard error, around line 414 of spm_graph.m), but I may have
> misinterpreted some of the calculations. Also even when plotting a
> single session (no averaging) the standard error is huge and the
> errors completely overlap from 2 different conditions in which a
> contrast shows a significant difference between them. [I would think
> the errors would not completely overlap if the fitted responses were
> significantly different from one another.]
The s.e. curves pertain to all the basis functions you used.
A t-contrast on just one basis function (eg the canonical hrf)
might therefore reveal a significant effect, even though the
s.e. curves overlap to some extent. (Nonetheless, the overlap
should normally be small at some PST if there is a highly
significant difference - ie the fitted response of one condition
should lie outside the s.e. of the other condition). If you
only used one basis function (the once tested in the contrast),
then it does sound fishy, and perhaps you could tell us more.
Rik
---------------------------8-{)}-------------------------
DR R HENSON EMAIL [log in to unmask]
Wellcome Department of
Cognitive Neurology TEL (work1) +44 (0)20 7833 7483
12 Queen Square TEL (work2) +44 (0)20 7833 7472
London, WC1N 3BG FAX +44 (0)20 7813 1420
URL: http://www.psychol.ucl.ac.uk/rik.henson/index.html
---------------------------------------------------------
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|