

Negative residual variance and correl... 

Message/Author 

Ed Fryer posted on Thursday, May 18, 2006  8:37 am



Hello. I’ve attached a printout with the basic problem we’re having. We want to use structural equation modeling on a child abuse dataset. The type of analysis is COMPLEX, and we are using strata, psus, weighting, and specifying subpopulations. Our major problem is negative residual variance and correlations greater than 1, which I believe constitute the Haywood case. I believe you can set the residual variance to 0 or some very small positive value to correct this, but that cannot be done withTYPE=COMPLEX. Still, we would like to run the procedure with TYPE=GENERAL, or another option just to get an approximation that does not have national inferences, but still would inform. I do not know the syntax for the MODEL CONSTRAINT option. Could you possibly look at the attached and tell me how to do that for F4 which is my principal outcome in this application? Another problem we are encountering often is lack of convergence. I assume this is just due to model lack of fit, since I set the iterations and convergence in an effort to overcome the problem. Again, thanks so much for any help you might be able to give. We are most grateful. Ed Mplus VERSION 4.0 MUTHEN & MUTHEN 05/17/2006 1:34 PM INPUT INSTRUCTIONS TITLE: MODEL FOR CHILDREN 10+ YEARS OF AGE; DATA: FILE IS 'C:\0\NSCAW\MPLUSF2'; FORMAT IS F6 F1 F2 F2 F2 F2 F2 F2 F3 F3 F3 F3 F3 F4 F5 F5 F5 F2 F2 F2 F2 F2 F7.4 F3 F2 F2 F2 F2 F3 F3 F3 F3 F3 F2 F2 F3 F3 F3 F2; VARIABLE: NAMES ARE nscawid stratum nscawpsu chdagey yrp_tot yyb_tpt nts_ipt nts_ept yah_age yah_r_ss yah_r_ws yah_m_ss yah_m_ws nanalwt nanalwt2 nanalwt3 nanalwt4 zcsernum zservnum ppcgpr fampr compr1 dperooh ycd_tpt rp3_tot cd3_tpt ts3_ipt ts3_ept ah3_age ah3_r_ss ah3_r_ws ah3_m_ss ah3_m_ws yb4_tpt yb3_tpt so1 so3 so4 tserv; USEVARIABLES ARE ppcgpr, fampr, compr1, so1, ycd_tpt, yah_r_ss, tserv, zservnum, dperooh, so3, cd3_tpt, ah3_r_ss; SUBPOPULATION IS chdagey>5 and chdagey<19; MISSING IS BLANK; WEIGHT IS nanalwt3; STRATIFICATION IS stratum; CLUSTER IS nscawpsu; ANALYSIS: TYPE=COMPLEX; ITERATIONS = 100000; CONVERGENCE = .0000001; MODEL: f1 by ppcgpr, fampr, compr1; f2 by so1, ycd_tpt, yah_r_ss; f3 by dperooh, tserv, zservnum; f4 by so3, cd3_tpt, ah3_r_ss; f3 on f1; f4 on f3 f2; OUTPUT: STANDARDIZED SAMPSTAT; 


I need more information to answer your question. Please send your input, data, output, and license number to support@statmodel.com. 

Back to top 

