

Distrubution of Est./S.E. in WLS, WLS... 

Message/Author 

daniel posted on Thursday, February 13, 2003  6:39 am



When indicators are all categorical or part of indicators are categorical, the estimators can be used in Mplus are WLS, WLSM, and WLSMV. In output there is Est./S.E. for each estimate. My question is what is (asymptotic) distribution of Est./S.E.? Can it be treated as standard normal or students tdistribution? If not, how to use that value to judge the significance of the estimate? Could you give me any reference? 


Standard normal. See the following reference: Muthén, B. & Satorra, A. (1995). Technical aspects of Muthén's LISCOMP approach to estimation of latent variable relations with a comprehensive measurement model. Psychometrika, 60, 489503. 

Heike B. posted on Tuesday, December 13, 2011  4:13 am



Hello, I am building a manifest path model with categorical variables using WLSMV, the variables had been skewed. Can I derive a onetailed pvalue from the twotailed pvalues given in the MODEL RESULT section by deviding the twotailed pvalue through 2? Thanks a lot in advance. Heike 


You will need to look up the value of the ratio of the parameter estimate to its standard error in a ztable to obtain the pvalue. 

Back to top 

