Full-MLE and MPlus estimator in the p... PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Sanjoy posted on Thursday, June 30, 2005 - 3:43 am
Prof. Muthen ...I asked you the question before and you said "yes" as an answer, however, till date I couldn't comprehend how would it be possible and the reasons behind are twofold … can you kindly explain it once more (let me first state the model that I'm estimating)

B by B1-B3;
R by R1-R3;
U on R B X1;
R on B X2;
B on R X3;

(where U is 0/1, Bi's and Ri's are 1/2/3/4/5 and Xi's share some common element)

Problem no 1.) How can we fit Full-information ML to this model given the fact the underlying theory assumes normality and in MPlus ML is defined ONLY for Logit and NOT for Probit framework ...(I'm referring to the Likelihood function from Browne & Arminger, 1995, chapter 4, Equation 4.120 and 4.121) …plus I was reading an article by Hahn and Soyer on Multivarite Logit and Probit comparison, from there article it looks that the similarity of results from Logit and Probit hold better for univariate rather than the multivariate case

Problem no 2.) in Fact I have tried to run this model using ML estimator ...it could NOT run as long as you have non-recursivity in the model

using WLSMV I got satisfactory result, however I want to take the advantages of numerical integration routine that Mplus has …

Thanks and regards
 bmuthen posted on Sunday, July 03, 2005 - 4:35 pm
1. With ML, normality is assumed for latent variables, whereas logit is postulated for regressions with dependent observed categorical variables.

2. It seem like this should work fine. Send your input, output and data to support@statmodel.com
 Sanjoy posted on Sunday, July 03, 2005 - 6:18 pm
Ok Professor ... thank you, I will
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: