|
|
Covariance matrix - toplix or homogen... |
|
Message/Author |
|
|
Is it possible to make the covariance matrix homogeneous or toplix as it is in HLM for multilevel longitudinal analyses? I see that you can check the autocorrelation in example 6.17, but are there any other options for changing the covariance matrix? Thanks Matt |
|
|
You can allow any identified covariances of the residuals over time using WITH and you can use any functions of such covariances (and variances) that you can describe by Model Constraint. I think this includes the homogeneous case and the Toeplitz case. I don't think we can handle cases with rho^{|t - t'|} for individually-varying times t, t'. |
|
|
I should also mention the equality constraint possibility that you have in the Model command, e.g. holding residual variances equal over time, or lag-1 correlations equal over time. |
|
Back to top |
|
|