 |
 |
Dear spmers, we were wondering if there were someone who could response this mail found in SPM mail-list (see bellow) since we are interested in the answer. Many thanks Carles Falcon IDIBAPS-Hospital Clinic Barcelona Unidad de Biofísica. University of Barcelona %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% item #026807 Subject: Re: Using temporal derivative From: Thilo Kellermann <[log in to unmask]> Reply-To: Thilo Kellermann <[log in to unmask]> Date: Wed, 24 Jan 2007 14:06:43 +0100 Content-Type: text/plain Dear Vince, I have basically two questions w.r.t. the use of the temporal derivative as you propose in your paper. As you wrote, you calculated a single image quantity per subject (and condition) and then performed t-tests on the second level, probably using an equation similar to equation 10 in your paper. 1. Is my guess correct to calculate these single image quantities not using equation 10 exactly, but rather the term in the numerator (i.e. sign(beta_1) * sqrt(beta_1^2 + beta_2^2) ? So formula 10 would calculate something like spmT-images on the first level, and omitting the residual error in the numerator will give us the first-level con-images, that can be taken to the second level (in case I am right?!?). 2. In equation 4 you mention something about normalization and orthogonalization of the time course. It is not quite clear to me if these procedures are really necessary for calculating the single images per subject. Do you refer to normalization of the data or of the model regressors (I would assume the latter...)? If this normalization and orthogonalization is necessary, would you be so kind to share the (few lines of??) Matlab-code to perform these? At first and second sight these procedures are not that straightforward to me...... Sorry for that! Thank you very much !! Best regards, Thilo ------------------------------------------------- This mail sent through IMP: http://horde.org/imp/