Dear Madi,
well, as I wrote in my first reply, this is not an easy issue. The
problem of task-correlated movements still awaits its proper solution.
The covarying out of variance is a practical way of dealing with a
problem for which we don't really have complete solution. However, it
does not work properly for task-correlated movements. I guess at the
moment the "scientifically correct" way to handle data sets with
task-correlated movements would be to through them away. We don't want
to do that, so a reasonable way of putting these data to some use would
be to analyse them with and without this correction while being aware
of the potential errors arising from each of the approaches. If that
strategy yields interpretable results those could be reported, albeit
with a very clear recognition of the problems involved. If it does not
yield interpretable results, I don't really see much hope for the
data.
I am sorry to sound a bit defeatist, and I am sure that some day there
will be a solution to task-correlated movements, but as I see it it
just isn't there yet.
Good luck Jesper
> Hello Jesper again;
>
> >>Hence, my recommendation to you would in SPM96 jargon read as: Try to
> analyse your data with and without the adjustment, which in your case
> is first order adjustment with spin history.
>
>
> I tried that already. The global effect i'm trying to measure varies
> depending on the alignment procedure, and for some subjects the results
> are opposite. The local effect does not.
>
> The litreature has contradictory opinions about this. My guess is that
> part of it is related to alignment. So simply, i'm confused.
>
> In simple terms, Is it scientifically correct to sampling errors, if
> this leads to contradictory results???
>
> Regards;
>
> -s madi
>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|