Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: need help to perform SEM for observed continuous and categorical data

From   William Buchanan <>
To   "" <>
Subject   Re: st: need help to perform SEM for observed continuous and categorical data
Date   Sat, 8 Jun 2013 15:25:35 -0700

Your question isn't entirely clear.  The SEM capabilities of Stata do not currently extend to mixture models and/or latent class models, so estimating a categorical or dichotomous latent variable wouldn't be possible that way (although there is likely a way to do it using the user written command - gllamm -).  If your covariates/independent variables are categorical, transform them into indicator variables first (the only issues are with categorical/dichotomous variables as dependent variables).  

There isn't any way to give you any better feedback regarding why the - polychoric - command (written by Stas Kolenikov and available via his personal site) cannot find the derivatives without knowing more about the data on hand.  However, if your data contain missing values in the variables of interest that may be a place to start.  

In response to your second question, you should probably read through the SEM manual a little bit.  Fitting the model from summary statistics is covered extensively in the manual and in the help file - h sem ssd -. 


Sent from my iPhone

On Jun 8, 2013, at 12:59, nadiahassan <> wrote:

> Hi, 
> I am working on SEM model that has more than50 observed variables. these
> include continuous observed variables for dependent variable whereas,
> continuous, dichotomous and categorical observed variables for independent
> latent variables.
> I read on statlist archives to use polychoric command for performing sem on
> categorical data. I have installed the command. 
> My questions are:
> Q1: Do I have to use this command for all 50 vars at a tiem? I tried it but
> result window displays 
> " could not calculate numerical derivatives
> missing values encountered"
> Q2: If i estimate polychoric matrix how can i do sem on this info?
> Regards
> --
> View this message in context:
> Sent from the Statalist mailing list archive at
> *
> *   For searches and help try:
> *
> *
> *

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index