I am attempting to use the RWLS toolbox for my 1st level analysis, but I am running into an issue of singularity with the model estimation that results in the inability to examine parameter contrasts (Results). The data, which is successfully entered into the design matrix, can be processed using the typical SPM AR(1) covariance structure, however when choosing the weighted least squares option, I get the following warnings during the Model Estimation (ellipses denote redundant content) resulting in a failure of the Contrast Manager:
Running 'rWLS Model estimation’
SPM8: spm_rwls_spm (v$Rev: 3.0 $) 22:46:27 - 04/09/2012
=========================================================
Initialising parameters : ...done
Plane 68/68 , block 1/1 : ...done
Temporal non-sphericity (over voxels) : ...ReML estimation
ReML Block 1
ReML Iteration : 1 ...5.214427e+01
.
.
ReML Block 3
ReML Iteration : 1 ...4.219083e+02
Warning: Matrix is close to singular or badly scaled.
Results may be inaccurate. RCOND = 2.785926e-56.
> In spm_rwls_reml at 136
In spm_rwls_spm at 911
In spm_rwls_run_fmri_est at 44
In matlabbatch/private/cfg_run_cm at 29
In cfg_util>local_runcj at 1477
In cfg_util at 803
In cfg_ui>MenuFileRun_Callback at 1218
In gui_mainfcn at 96
In cfg_ui at 53
.
.
Done
Done 'rWLS Model estimation’
Running 'Contrast Manager'
1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 <- !invalid contrast
============================================================
The invalid contrast is a result of the design matrix having estimatable parameters (betas uniquely specified) prior to RWLS model estimation to having unestimatable parameters (betas not uniquely specified) afterwards.
Does anyone have any ideas as to why this might occur or how to fix/bypass the problem?
|