About the output of lavaan for lisrel model

asas asas

unread,

Jan 15, 2019, 5:43:43 PM1/15/19

to lavaan

Dear All,

I have a question about the lavaan package for the general SEM or lisrel model.There is a parameter matrix $Phi$,which makes up the covariance matrix of the exogenuous latent variables and is to be estimated.

Does lavaan check whether the estimated matrix of this $Phi$ is also positive definite?Or the elements of this matrix are estimated freely.

In the file 'lavaan_model_estimate.R',it seems to check the positive definiteness of the initial value of Sigma.However this Sigma may not be the $Phi$, the covariance matrix of the exogenuous latent variable.

Thanks,

Tom

Terrence Jorgensen

unread,

Jan 16, 2019, 6:11:09 PM1/16/19

to lavaan

There is a parameter matrix $Phi$,which makes up the covariance matrix of the exogenuous latent variables and is to be estimated.

lavaan only uses the "Y side" of the LISREL model. So there is no Phi, only Psi, which includes (co)variances of all (exogenous and endogenous) latent variables.

Does lavaan check whether the estimated matrix of this $Phi$ is also positive definite?

Yes, if the the "check.post" option is set to TRUE, it will check that after estimation.

lavOptions("check.post")

In the file 'lavaan_model_estimate.R',it seems to check the positive definiteness of the initial value of Sigma.

Yes, it checks the model-implied Sigma *implied by the starting values" before estimation, if the "check.start" option is set to TRUE.

lavOptions("check.start")

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

http://www.uva.nl/profile/t.d.jorgensen

asas asas

unread,

Jan 22, 2019, 4:02:03 PM1/22/19

to lavaan

Thanks.

A final extra question is that, for those variances of the residuals,does lavaan parameter them in the scale of sigma^2 or sigma?If sigma^2 is parameterized, does lavaan add the constraints that they should be positive?

Best,

Yilun

在 2019年1月16日星期三 UTC-5下午6:11:09，Terrence Jorgensen写道：

Terrence Jorgensen

unread,

Jan 25, 2019, 9:52:31 AM1/25/19

to lavaan

does lavaan parameter them in the scale of sigma^2

Yes, they are variances, not SDs.

If sigma^2 is parameterized, does lavaan add the constraints that they should be positive?