Rik -
Apologies for the delay, but I've been away.
> 1. When averaging over sessions, the s.e. appears to be really large in
> comparison to the magnitude of the fitted response even for highly
> significantly activated voxels (see attachment). Needless to say it becomes
> smaller when we divide s.e. by the sqrt of the total number of sessions but
> there may be a better explanation?
The s.e. of the FITTED RESPONSE is not calculated correctly in the code
provided. The code was simply to plot the average fitted response over
sessions. The appropriate s.e. for the average fitted response is not
clear. You could simply take the average s.e. of the session-specific
fitted responses by inserting the line after line 418:
se = se*ones(length(ss))/length(ss);
but I'm not sure this is particularly meaningful (it is not the s.e.
had the same mean response been fitted for all sessions).
The mean and s.e. of the ADJUSTED DATA (PSTH) are "correct", but only
in the sense of a fixed effects model. In the devel code provided, the
averaging over sessions is performed simply by concatinating the data
from each session, then taking the grand mean and grand s.e. The s.e.
here is not equivalent to the s.e. of the means for each session (i.e.
within- and between-session variability are confounded). If you want
the s.e. to pertain to that over sessions (or subjects) only, you can
insert code to perform an averaging within-session first (quite easy to
do).
> 2. When plotting event-related responses averaged over sessions, s.e. is
> plotted as several curves, do these correspond to confidence intervals?
This is an error. Each curve is a session-specific s.e.
> 3. At fixed time points, e.g. for event-related at t=0 and t=12 s, s.e.
> becomes zero. Why?
If you are referring to the adjusted data, this can arise if there is
only one data point for those time-bins.
> 4. In an event-related study, what does it mean if at t=0 (event onset) the
> fitted response happens not to be 0?
I expect you temporally smoothed, in which case this is a consequence of
the smoothing.
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
---------------------------------------------------------
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|