EFA -No Convergence: number of iterat...
Message/Author
 Kristin Tomey posted on Friday, June 21, 2013 - 10:03 am
Hi there,

I have an error message from my EFA output:

NO CONVERGENCE. NUMBER OF ITERATIONS EXCEEDED. PROBLEM OCCURRED IN EXPLORATORY FACTOR ANALYSIS WITH 10 FACTOR(S).

The User Guide specifies that for this type of convergence problem, I should check to make sure there are no large negative variances / residual variances in the prelim parameter estimates.

So for each section of the output (EFA with 2 factors, EFA with 3 factors, etc etc) I have looked at the "estimated residual variances". I don't know what "large" would be. Can you help me understand? Is a residual variance of 0.5 large? 0.6? 0.9?

Thank you,
KT Tomey
 Linda K. Muthen posted on Friday, June 21, 2013 - 3:24 pm
Please send the output and your license number to support@statmodel.com.
 Andriana Rapti posted on Monday, December 09, 2013 - 7:12 am
I get this error message in my EFA:

NO CONVERGENCE IN ROTATION ALGORITHM. NUMBER OF ITERATIONS EXCEEDED.
PROBLEM OCCURRED IN EXPLORATORY FACTOR ANALYSIS WITH 2 FACTOR(S).
THE PROBLEM IS PROBABLY DUE TO THE NEGATIVE RESIDUAL VARIANCE FOR
VARIABLE F6.

However, when I switch to estimator = bayes,it is performed. Is this correct?
 Linda K. Muthen posted on Monday, December 09, 2013 - 10:51 am
Bayes keeps residual variances positive through the use of priors. I would consider the one-factor solution.
 Andriana Rapti posted on Monday, December 09, 2013 - 5:07 pm
But I factor gives me significant chi-square and high RMSEA. What should I do? Should I just keep these values?
 Andriana Rapti posted on Monday, December 09, 2013 - 5:50 pm
I am sorry, I meant one-factor gives me significant chi-square and high RMSEA...
 Linda K. Muthen posted on Tuesday, December 10, 2013 - 11:14 am
You can consider not using that item. If the residual variance is very small and not significant, you would do EFA via ESEM and fix it at zero.
 Andriana Rapti posted on Tuesday, December 10, 2013 - 2:15 pm
I tried it so I added in my model command:
F BY F1 F2 F3 F4 F5 (*1);

(instead of F BY F1 F2 F3 F4 F5;), but I am getting the same results and the same values. What am I doing wrong?
 Linda K. Muthen posted on Tuesday, December 10, 2013 - 5:03 pm
Please send the output that shows the negative residual variance and your license number to support@statmodel.com.
 DEC posted on Tuesday, May 31, 2016 - 12:51 pm
Hello,

I ran an EFA (1 - 5) on my data and Mplus conveniently skips the output of the '3 factor model'. I've tried this on various Mplus versions and I'm still getting the same blank result. Can you please help me produce output showing all factor models?

Thank you!

Daniel
 Linda K. Muthen posted on Tuesday, May 31, 2016 - 1:12 pm
Please send the output and your license number to support@statmodel.com.
 JuliaSchmid posted on Monday, January 30, 2017 - 10:55 pm
Dear Mplus-Team

I wanted to run an ESEM (with correlated residuals), but I didn't suceed. Following error message occurred: "NO CONVERGENCE. NUMBER OF ITERATIONS EXCEEDED. THE CHI-SQUARE STATISTIC IS NEGATIVE.THE LOGLIKELIHOOD VALUES MAY NOT BE RELIABLE." Do you have any advice what I could do?

Thank you in anticipation!

Input:

ANALYSIS: ESTIMATOR = MLR;
ROTATION=TARGET;

MODEL:
f1 BY kon1 kon2 kon3 kon4 kon5
wetlei1~0 wetlei2~0 wetlei3~0 wetlei4~0
ablkat1~0 ablkat2~0 ablkat3~0 ablkat4~0
figaus1~0 figaus2~0 figaus3~0
fitges1~0 fitges2~0 fitges3~0
aktfre1~0 aktfre2~0 aktfre3~0
aes1~0 aes2~0(*1);

f2 BY kon1~0 kon2~0 kon3~0 kon4~0 kon5~0
wetlei1 wetlei2 wetlei3 wetlei4
ablkat1~0 ablkat2~0 ablkat3~0 ablkat4~0
figaus1~0 figaus2~0 figaus3~0
fitges1~0 fitges2~0 fitges3~0
aktfre1~0 aktfre2~0 aktfre3~0
aes1~0 aes2~0(*1);

...

f7 BY kon1~0 kon2~0 kon3~0 kon4~0 kon5~0
wetlei1~0 wetlei2~0 wetlei3~0 wetlei4~0
ablkat1~0 ablkat2~0 ablkat3~0 ablkat4~0
figaus1~0 figaus2~0 figaus3~0
fitges1~0 fitges2~0 fitges3~0
aktfre1~0 aktfre2~0 aktfre3~0
aes1 aes2(*1);

kon4 WITH kon5;
 Linda K. Muthen posted on Tuesday, January 31, 2017 - 12:12 pm
Please send the output and your license number to support@statmodel.com.
 Ashley Duguay posted on Thursday, April 19, 2018 - 8:54 am
Dear Mplus Team,

Similar to Julia above, I tried to run an ESEM but received the following warnings:

NO CONVERGENCE. NUMBER OF ITERATIONS EXCEEDED.

WARNING: THE RESIDUAL COVARIANCE MATRIX (THETA) IS NOT POSITIVE DEFINITE. THIS COULD INDICATE A NEGATIVE VARIANCE/RESIDUAL VARIANCE FOR AN OBSERVED VARIABLE, A CORRELATION GREATER OR EQUAL TO ONE BETWEEN TWO OBSERVED VARIABLES, OR A LINEAR DEPENDENCY AMONG MORE THAN TWO OBSERVED VARIABLES. CHECK THE RESULTS SECTION FOR MORE INFORMATION. PROBLEM INVOLVING VARIABLE UTOWE5.

THE CHI-SQUARE STATISTIC IS NEGATIVE. THE LOGLIKELIHOOD VALUES MAY NOT BE RELIABLE.

No model fit indices were provided in the output, outliers were not computed, and no data were saved.

Extra information: I used TYPE=CLUSTER to account for nested data, ESTIMATOR=MLR, and ROTATION=TARGET.

Do you have any advice on how I may be able to proceed?

Thank you in advance.
 Bengt O. Muthen posted on Thursday, April 19, 2018 - 4:12 pm
Please send to Support along with your data and license number.
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: