CFI/TLI > 1 PreviousNext
Mplus Discussion > Exploratory Factor Analysis >
Message/Author
 anon9210 posted on Friday, September 17, 2010 - 12:43 pm
Is it a problem if CFI and TLI are greater than 1 in an EFA? Or if RMSEA is exactly 0?

There don't seem to be any other issues with the EFA model - i.e., residual variances are all +ve, loadings look as expected, etc.

Thanks in advance again!
 Linda K. Muthen posted on Friday, September 17, 2010 - 2:46 pm
Please send your output and license number to support@statmodel.com.
 Miranda Vervoort posted on Monday, October 11, 2010 - 6:40 am
I have the same problem with my path models. CFI is exactly 1 and the TLI is greater than 1.
What can I do about this?

Thank you in advance!
 Linda K. Muthen posted on Monday, October 11, 2010 - 6:52 am
TLI can be greater than one. It sounds like your model is just-identified with zero degrees of freedom. In this case, model fit cannot be assessed.
 Miranda Vervoort posted on Wednesday, October 13, 2010 - 12:40 am
I don't think that is the problem. The chi-square test of model fit indicates a value of 31.339 with 33 degrees of freedom and the number of free parameters is 64.
But does these high values of CFI/TLI indicate that there are problems with my model?
I don't get any errors and the findings are highly in line with the theoretical expectations.
 Linda K. Muthen posted on Wednesday, October 13, 2010 - 11:37 am
From what you say the p-value for chi-square must be very large which like CFI and TLI points to a well-fitting model.
 Michaela Sokoliová posted on Tuesday, April 03, 2012 - 3:04 am
Hello ;)
I would desperately need your help and appreciate any of your suggestions ;)
So...
I used EFA (after running CFA to confirm 2factors structure of my guestionnaire – that was rejected) for my data (N = 213, items 11) and I got:
3 factors: df = 25; x2 = 19.890; p = .7525; CFI = 1.000; TLI = 1.022; RMSEA = .000
4 factors: df = 17; x2 = 9.730; p = .9145; CFI = 1.000; TLI = 1.046; RMSEA = .000

3 factors model seems to be more appropriate to me but... I have no clue how to report this (if I reject the model with 4 factors). Both models should be acceptable according to statistics or??? I´ve already read, that´s no problem if TLI is higher than 1.

I´m really thankful in advance 
 Linda K. Muthen posted on Tuesday, April 03, 2012 - 10:05 am
Once you get a p-value greater than .05 you should stop. So three factors would seems best. You should also use theory to support this decision.
 Michaela Sokoliová posted on Tuesday, April 03, 2012 - 11:11 am
Thanks a million ;)
Yeah, theory is supportive but I wanted to hint as statistic reason as well. So if this is enough I´m fine with it.
thank you again
 RGL posted on Wednesday, June 13, 2012 - 2:28 pm
Hello,

I am in a similar situation, after an EFA with binary variables: RMSEA = 0.000, CFI and TLI greater than 1.00. Chi-squared stats are good, and the fit works theoretically. From this thread, it sounds like this is fine, but I'd love a reference if you have one!

Thanks!
 Linda K. Muthen posted on Wednesday, June 13, 2012 - 3:03 pm
Please send the output and your license number to support@statmodel.com. CFI cannot be greater than one.
 RGL posted on Wednesday, June 13, 2012 - 3:22 pm
Hello - sorry, my mistake, CFI = 1.00, TLI slightly larger than 1.00.

I am using MPlus at a computer lab at the University of Michigan (where I am a full-time student), not sure about obtaining the license number
Thanks!
 Stephen Teo posted on Monday, February 04, 2013 - 5:03 pm
Hello

My CFI = 1.000 TFI = 1.004
My Chi-sq = 1.939, df=3 and p = 0.5851
SRMR=0.009

Is there an issue with this model?

thanks!
 Bengt O. Muthen posted on Monday, February 04, 2013 - 6:08 pm
You probably have low correlations and/or low sample size so not much power to reject

TFI >1 should be rounded to 1.
 kja posted on Monday, May 06, 2013 - 6:15 pm
Hello Dr. Muthen,

Is it concerning to have a CFI=1.00 and RMSEA=0.00? (I am using MLR). It does not appear that there are any issues with model identification. I ran the same model in EQS, and while the standardized path betas and factor loadings were nearly equivalent, the CFI=0.991 and RMSEA=0.023.

Thank you.
 kja posted on Monday, May 06, 2013 - 6:19 pm
Hi Dr. Muthen,

Sorry - please disregard the previous post, I figured out the issue. Thank you.
 Stefano Cacciamani posted on Thursday, December 14, 2017 - 12:41 am
Hi Dr Muthen,
I'm analyzing a model with 4 variables in a sample of 93 subjects. The MPLUS output is the following ( see below) with CFI and TLI =1.
What does it means? Is it a problem the sample size?
Thanks in advance for your kind attention.
Stefano Cacciamani

Chi-Square Test of Model Fit

Value 0.000
Degrees of Freedom 0
P-Value 0.0000

RMSEA (Root Mean Square Error Of Approximation)

Estimate 0.000
90 Percent C.I. 0.000 0.000
Probability RMSEA <= .05 0.000

CFI/TLI

CFI 1.000
TLI 1.000

Chi-Square Test of Model Fit for the Baseline Model

Value 25.458
Degrees of Freedom 9
P-Value 0.0025

SRMR (Standardized Root Mean Square Residual)

Value 0.000
 Stefano Cacciamani posted on Thursday, December 14, 2017 - 12:47 am
Dear dr. Muthen,
just a correction about the previous post: the variables are 5 instead of 4.
Thanks again
Stefano Cacciamani
 Bengt O. Muthen posted on Thursday, December 14, 2017 - 12:27 pm
See our FAQ:

TFI fit index: Unusual values
 Es Maths posted on Wednesday, February 12, 2020 - 3:42 pm
Hello, should I be concerned with RMSEA = 0, TLI = 1.008, CFI = 1, and p = .570. This is from a latent growth model in which I also fixed the residual variance of first time point to zero.

model:

i s | CK1@0 CK2@1 CK3@2;

CK1@0;
 Bengt O. Muthen posted on Wednesday, February 12, 2020 - 4:46 pm
Sounds like a very well-fitting model. Which can be due to small sample size and small correlations.
 Amelia Rock posted on Monday, April 20, 2020 - 1:30 pm
Hello Dr. Muthen,
When you say "small correlations" in the last post on this thread, do you mean correlations amongst independent variables, or between independent and dependent variables? Many Thanks
 Bengt O. Muthen posted on Tuesday, April 21, 2020 - 1:13 pm
Both. But you seem to have only DVs.
 Amelia Rock posted on Wednesday, April 22, 2020 - 7:46 am
Many thanks for the response. The posting to which I referred was actually not mine. My model is a mediation model with 10 Xs, one continuous mediator, and one binary outcome (and controls). I used probit link function and robust WLS estimator. Sample size is 223. The model fit info is:
Chi-Sq = 3.842, 6 DOF, p= 0.6980
RMSEA = 0.000, p=0.888
CFI = 1.000
TLI = 10.043

If you have any thoughts on whether I should be concerned about the CFI, TLI, and RMSEA values, I would be grateful. For reporting, I plan to round the TLI to 1 per your guidance re: a different post. Thanks again.
 Tihomir Asparouhov posted on Wednesday, April 22, 2020 - 10:56 am
In the current version 8.4 it is already reported as 1. All results are consistent with each other so I don't think there is any concern.
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: