Scott -
> I am currently attempting to understand the different plotting functions in
> SPM99, and have plotted graphs for the same voxel using different contrasts.
> When I choose "event & epoch related responses" and then "fitted response and
> PSTH", and choose the same trial number, the same graph is output regardless of
> what contrast I have selected.
>
> First, I would like to confirm that the waveforms w/ dashed lines represent the
> fitted response and the waveforms with error bars represent the PSTH.
Yes.
> Second, is it correct to conclude that I am graphing the mean bold response of
> that voxel for all instances of that condition, irrespective of contrast, and
> that there is no "baselining." That is, activation from the control condition
> has not been subtracted from the activation of the experimental condition. Or
> put another way, the amplitude of the waveform for the control conditon has not
> been subtracted from the amplitude of the waveform for the experimental
> condition.
You are correct - there is no "baselining" - assuming that you modelled
the control condition too.
(If the control condition were unmodelled - and were present for all
scans in which the experimental condition was absent - then it would
become the baseline and hence "subtracted").
> Third, if the above is true (no differences in waveform using different
> contrasts and no baselining), and I wanted to plot the amplitude of the
> waveforms of the experimental conditions minus the amplitude of the control
> condition, how would I do that? Does SPM have this capability? If not, is
> there a way to output the numerical values (the PSTH) for each timepoint in
> each condition so that I could make the graphs and do the statistics myself w/
> other software?
SPM does not plot contrasts of event-related responses unfortunately.
If you have used a single basis function, then an estimate of the
difference between experimental and control conditions would be
perfectly well summarised by the difference in the parameter estimates
(which you can plot contrasts of via the GUI "plot" menu).
If you have more than one basis function, and want to characterise
the difference in terms of single number, then you might calculate
the difference in peak of the fitted responses, but you must remember
that this quantity is not actually the quantity about which you made
a statistical inference. It would be more appropriate to plot the
"experimental - control" differences in each basis function (by plotting
an F-contrast comprising [1... -1...] for each basis function; eg
[eye(N) -eye(N)] for N basis funcitons). Once you realise that the
PSTH is NO MORE THAN the set of parameter estimates for an FIR basis
set, you will see the generality of this approach. Thus you could
refit the model using an FIR basis set and plot F-contrasts to see
the form of the differential event-related response (but again, your
inference should really also be based on this new model).
Rik
--
---------------------------------------------------------
DR RICHARD HENSON
Institute of Cognitive Neuroscience
& Wellcome Department of Imaging Neuroscience
University College London
17 Queen Square
London, WC1N 3AR
England
EMAIL: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~rhenson
TEL1 +44 (0)20 7679 1120
TEL2 +44 (0)20 7833 7472
FAX +44 (0)20 7813 1420
MOB +44 (0)794 1377 345
---------------------------------------------------------
--
|