Dear Rik and other statistics experts,
Are there good literature references as to why T-tests are not allowed when
the HRF is modelled by [more than 1] basis functions? What exactly goes
wrong if you just compute [beta1 + beta2 +...]/[ResMS]?
For 1 basis function per HRF, both tests select the same set of points.
regards
Alle Meije
>> I have an event-related analysis with the canonical hrf and its two
>> derivatives. Most of the variance is explained by the time derivative of
>> the hrf, so I would like to make a contrast in which the time derivative
>> has a higher weight than the hrf and the dispersion derivative.
>
>You should not be using T-tests that span more than one basis function;
>you should use F-tests for multiple basis functions. And I don't think
>it makes much sense to give more weight to one basis function relative
>to others.
>
>Rik
>
>DR RICHARD HENSON
>Institute of Cognitive Neuroscience
>& Wellcome Department of Imaging Neuroscience
>University College London
>17 Queen Square
>London, WC1N 3AR
>England
|