depsampregrT: no difference between contrasts?

[Jonas Obleser]

Hello FTs,
here comes a question concerning the wonderful stats function

depsampregrT

I am using it mainly for running time–frequency–electrode cluster tests. I usually use contrasts like cfg.contrastcoefs=[-3 -1 1 3] and it has proven pretty powerful.

With a new data set, I sadly have only three levels of data, which is of course always a bit of a pain for linear contrasts.
What puzzles me is that

[-2 -1 3] (i.e., linear increase; trying to keep it zero-weighted)

and 

[-1 2 -1]  (i.e. trying to crudely model an inverse u-shape relationship)

give the very same result (not exactly numerically,  but very much in terms of clusters, down to very similar p-values, etc).

Q: Any idea why this might be the case?
Of course, linear and cubic trends can be sometimes both significant, but I was slightly worried seeing this strong concordance in this case.

(Also, does any of you have a better/best/gold standard way to model 3-level linear contrasts? mathematically, in [-1 0 1], the middle condition is basically switched off, isn’t it.)

Anyway, thanks very much for a quick reply!
Best wishes, Jonas




http://jonasobleser.com
+49 171 6993337 mobile
+49 341 304 7980  home






----------------------------------
The aim of this list is to facilitate the discussion between users of the FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/archives/fieldtrip.html and http://www.ru.nl/neuroimaging/fieldtrip.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20100316/b35884f8/attachment-0001.html>