Hanno Petras posted on Thursday, September 15, 2005 - 6:13 pm
Dear Linda and Bengt,
I have run an EFA analysis using the self esteem items from the LSAY data. The problem I encountered that a one and three factor solution converged but not the 2 factor solution. I increased the number of interation and the 2 factor solution conseuqently converged. However, one of the items shows a loading of larger than 1 and has a large negative residual variance. Given that the same problem is also encountered in the 3 factor model, I am wondering if this shows that the last item is not useful and/or a one factor solution is preferrable. Any advice would be greatly appreciated. Below I have posten the converged output.
Mplus VERSION 3.13 MUTHEN & MUTHEN 09/15/2005 11:14 AM
Title: The variable names are for the data set lsay.dat
Analysis: Type is efa 1 3 missing; iterations=5000;
Output: sampstat patterns mod (3.84) tech1 ;
Plot: Type is plot1 plot2 plot3;
*** WARNING in Output command SAMPSTAT option for analysis types MISSING and MCOHORT requires H1. Analysis type H1 is turned on automatically. *** WARNING in Output command MODINDICES option is available only for Analysis types GENERAL and MIXTURE. Request for MODINDICES is ignored. *** WARNING Data set contains cases with missing on all variables. These cases were not included in the analysis. Number of cases with missing on all variables: 14 3 WARNING(S) FOUND IN THE INPUT INSTRUCTIONS
The variable names are for the data set lsay.dat
SUMMARY OF ANALYSIS
Number of groups 1 Number of observations 1476
Number of dependent variables 6 Number of independent variables 0 Number of continuous latent variables 0
Observed dependent variables
Continuous SELF WORTH OTHER SATISF RESPECT FAILURE
Variables with special functions
ID variable LSAYID
Estimator ML Information matrix OBSERVED Maximum number of iterations 5000 Convergence criterion 0.500D-04 Maximum number of steepest descent iterations 20 Maximum number of iterations for H1 2000 Convergence criterion for H1 0.100D-03
Input data file(s) j:\ccjs699\dataset\lsay.dat
Input data format FREE
SUMMARY OF DATA
Number of patterns 23
SUMMARY OF MISSING DATA PATTERNS
MISSING DATA PATTERNS
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 SELF x x x x x x x x x x x x x x x x x WORTH x x x x x x x x x x x x OTHER x x x x x x x x x x x x x x x SATISF x x x x x x x x x x x RESPECT x x x x x x x x x x x x FAILURE x x x x x x x x x x x x x
21 22 23 SELF WORTH OTHER x SATISF x x RESPECT x x FAILURE x
BMuthen posted on Thursday, September 15, 2005 - 8:32 pm
A negative residual variance may indicate overextraction of factors. This does not imply that one factor is sufficient. Perhaps the best model has one factor and some minor factors that can be captured by correlated residuals.
Please do not post full outputs in Mplus Discussion.
Tracy Witte posted on Monday, December 01, 2014 - 5:52 pm
I have a question about factor loadings larger than 1.0 when doing an EFA with ordinal variables, using the WLSMV estimator. In one of my factor solutions, one item has a geomin rotated loading of 1.203 on one of the factors. However, none of the items have negative residual variances. Additionally, there are no other warnings in my output, and the solution appears to have converged properly.
My question is this: do loadings greater than 1.0 with the WLSMV estimator in EFA necessarily constitute a Heywood case? Should I not be concerned since the residual variances are all positive?