Non Positive Definite PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
Message/Author
 Eun K Um posted on Monday, March 14, 2016 - 1:43 pm
I did confirmatory factor analysis by using the following input.

TITLE: CFA Trimmed Model 1
DATA: FILE IS C:\Mplus\Lewis.dat;
VARIABLE: NAMES ARE p d n c s a ne al ai ad se e l;
USEVARIABLES ARE p d n c s a ne al ai ad se e l;
MODEL: st BY p d n c;
dep BY s a;
sub BY ne al ai;
adh BY ad se;
adh ON st dep sub;
dep ON st;
sub ON st;
st dep sub adh ON e l;

But I got the following message:
WARNING: THE LATENT VARIABLE COVARIANCE MATRIX (PSI) IS NOT POSITIVE
DEFINITE. THIS COULD INDICATE A NEGATIVE VARIANCE/RESIDUAL VARIANCE FOR A LATENT VARIABLE, A CORRELATION GREATER OR EQUAL TO ONE BETWEEN TWO LATENT
VARIABLES, OR A LINEAR DEPENDENCY AMONG MORE THAN TWO LATENT VARIABLES. CHECK THE TECH4 OUTPUT FOR MORE INFORMATION.
PROBLEM INVOLVING VARIABLE ADH.

I don't understand exactly what does it mean and I don't know how to solve the problem. Could you help?
 Bengt O. Muthen posted on Monday, March 14, 2016 - 5:33 pm
Check TECH4 and look for what the message talks about.

This means that your model is not suitable for your data and you will have to modify the model. For instance, look at the Modindices.
 Eun K Um posted on Tuesday, March 15, 2016 - 10:36 am
Dr. Muthen, Thank you for your response. It is helpful. I added the Modindices as follows: I don't think I added it right way. I could not run this. In addition, how can I add TECH4?

TITLE: CFA Trimmed Model 1
DATA: FILE IS C:\Mplus\Lewis.dat;
VARIABLE: NAMES ARE p d n c s a ne al ai ad se e l;
USEVARIABLES ARE p d n c s a ne al ai ad se e l;
MODINCDICES(0);
MODEL: st BY p d n c;
dep BY s a;
sub BY ne al ai;
adh BY ad se;
adh ON st dep sub;
dep ON st;
sub ON st;
st dep sub adh ON e l;
 Bengt O. Muthen posted on Tuesday, March 15, 2016 - 12:12 pm
As you see in the UG the TECH4 request is done in the Output command. Modindices is misspelled.
 Eun K Um posted on Tuesday, March 15, 2016 - 12:20 pm
Dr. Muthen, Thank you for your response. It is helpful. I added the Modindices as follows: I got this message.

TITLE: CFA Trimmed Model 1
DATA: FILE IS C:\Mplus\Lewis.dat;
VARIABLE: NAMES ARE p d n c s a ne al ai ad se e l;
USEVARIABLES ARE p d n c s a ne al ai ad se e l;
MODINDICES(0);
MODEL: st BY p d n c;
dep BY s a;
sub BY ne al ai;
adh BY ad se;
adh ON st dep sub;
dep ON st;
sub ON st;
st dep sub adh ON e l;






*** ERROR in VARIABLE command
Unknown option:
MODINDICES(0)
 Linda K. Muthen posted on Tuesday, March 15, 2016 - 2:54 pm
You should put it in the OUTPUT command not in the VARIABLE command.
 Eun K Um posted on Wednesday, March 16, 2016 - 1:21 pm
Dear Dr. Muthen:

Thank you for your response. It is very helpful. I got the following output. Can you give me your advice on how I can change model based on the following output?

Minimum M.I. value for printing the modification index 0.000

ON/BY Statements

DEP ON SUB /
SUB BY DEP 0.259 2.196 0.042 0.042
DEP ON ADH /
ADH BY DEP 0.259 1.070 999.000 999.000
SUB ON DEP /
DEP BY SUB 0.259 0.001 0.052 0.052
SUB ON ADH /
ADH BY SUB 0.259 0.012 999.000 999.000
 Bengt O. Muthen posted on Wednesday, March 16, 2016 - 6:07 pm
This is a good general analysis question for SEMNET.
 Eun K Um posted on Thursday, March 17, 2016 - 1:42 pm
I deleted ADH and did not get the warning that the model is non positive definite. But my residual variances are high.
Please see the following:
TITLE: CFA Trimmed Model 1
DATA: FILE IS C:\Mplus\Lewis.dat;
VARIABLE: NAMES ARE p d n c s a ne al ai ad se e l;
USEVARIABLES ARE p d n c s a ne al ai e l;
MODEL: st BY p d n c;
dep BY s a;
sub BY ne al ai;
dep ON st;
sub ON st;
st dep sub ON e l;
OUTPUT:TECH4
Residual Variances
P 4.017 0.576 6.969 0.000
D 2.812 0.363 7.736 0.000
N 2.251 0.285 7.898 0.000
C 2.528 0.482 5.240 0.000
S 562.290 316.374 1.777 0.076
A 0.333 0.093 3.583 0.000
NE 0.222 0.050 4.460 0.000
AL 969.555 136.641 7.096 0.000
AI 881.951 147.108 5.995 0.000
ST 5.777 1.007 5.735 0.000
DEP 1503.962 345.716 4.350 0.000
SUB 0.671 0.102 6.607 0.000
 Bengt O. Muthen posted on Thursday, March 17, 2016 - 5:15 pm
For general analysis questions that don't relate specifically to Mplus capabilities, please try SEMNET.
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: