Hi Dorrit,
On 4 Apr 2012, at 09:32, Dorrit Inbar wrote:
> I am approaching you again, this time with another question that
> appeared during reading the manual.
> In the section of ERP analysis of EEG (page 361) it is suggested in
> the manual to perform a "2nd level" analysis of the images built
> from the data. I was wondering what is the reason for performing 2nd
> rather than 1st level analysis, and what could be an indication to
> perform a 1st level analysis.
> My question comes up since previously the text notes that 1st level
> is intended to within-subject analysis, whereas 2nd level is for
> between-subject. However, in our case in page 361 the analysis is of
> a singe subject. Therefore, what is the reason for performing the
> 2nd level analysis? Is it due to comparison between two trial
> conditions ("standard" vs "rare")?
> Is it only a matter of dimension size of the data- due to the time
> dimension within each trial? If so, is it always true we should
> perform 2nd level analysis on EEG data, comprised of more than a
> single bin per trial?
> Is there another difference in the statistical methodology between
> the 1st and 2nd levels, to be the reason for choosing 1st or 2nd
> level analysis?
> (Or, in other words, what am I missing here....?)
I think you are confused by some problematic terminology. In SPM8 all
the statistical analyses we do on images generated from MEEG data are
initiated via the 2nd-level button. The analysis in question can be
within subject (across trials) or across subjects. In all cases you
use the same statistical machinery, just possibly different designs.
For instance you can do regression within subjects followed by one-
sample t-test on beta images across subjects. In both cases you will
use the '2nd-level' button. '1st-level' button is only presently used
for averaging across time within an image to create single topography.
It is not really statistical analysis. In SPM fMRI 1st level button
invokes statistical analysis involving modeling the continuous BOLD
signal, so something not relevant for MEEG.
Best,
Vladimir
|