1. The values of beta_* will be the parameter estimate at each voxel.
2. They might be outside the mask or they might actually have 0 estimated effect.
3. You should have a constant term for each run. You should check the code for all of the steps. For example, there is often a pre-whitening step.
________________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on behalf of Philip Lin [[log in to unmask]]
Sent: Thursday, December 19, 2013 8:50 AM
To: [log in to unmask]
Subject: [SPM] Questions about (level 1 analysis) GLM in SPM
Hi, all.
I am a user of SPM8 and trying to understand the detail of level 1 GLM analysis in SPM.
I met some problems and hope someone can help me.
Q1. in the output of (level 1) beta estimation, there're several images (beta_0001.img, ..., beta_0020.img). Are these images the final parameter estimation of each covariates ? If I read one of these images into R, for example beta_0001.img, and see values in each voxel. Are the values I see exactly same as the parameter estimation ?
Q2. As the same example in Q1, I've found that not all voxels have estimated values (Most of voxels are 0, some regions have non zero values), why not all voxels were estimated ? or did these voxels have beta1_hat = 0 ?
Q3. In my experiments, I have two runs and each run with 3 regressors, 6 rigid body parameters and 1 constant. (thus, 2*(3+6+1)=20 parameters) Thus, the dimension of my design matrix is 320*20, denoted as X. Before GLM, my original voxels (y) and X would be multiplied by a high pass filter matrix (S = I - X0*X0'), and all estimation would be obtained against my filtered y and filtered X, am I right ? or are there other manipulations or transformations on y and X ?
Thanks for your time, very appreciate if there's any help.
Philip Lin.
|