Mplus Discussion >> Bootstrap and categorical data modelling

Topics
Last Day
Last 3 Days
Last Week
Tree View

Edit Profile


Bootstrap and categorical data modelling

Mplus Discussion > Categorical Data Modeling >

Message/Author

Francesca Pesola posted on Friday, May 04, 2012 - 7:11 am

Hi,

I am using SEM to explore a mediation model. The observed variables are binary/ordinal (while the latent variables are continuous).

I am using the default estimator: WLSMV.

I would like to use bootstrap on the data; but when I add analtsis: type = bootstrap, MPlus does not produce the model estimates. Well below is what I obtain from the output:

THE MODEL ESTIMATION TERMINATED NORMALLY

TESTS OF MODEL FIT

Number of Free Parameters 100

RMSEA (Root Mean Square Error Of Approximation)

Estimate 0.000

WRMR (Weighted Root Mean Square Residual)
Value 1.407

Can anyone tell me why I am receiving such input and whether it is possible to use the bootstrap option and still obtain the chi-square and CFI model fit indices?

Many thanks

Linda K. Muthen posted on Friday, May 04, 2012 - 9:57 am

TYPE = BOOTSTRAP does not exist in the Mplus language. It is BOOTSTRAP = the number of bootstrap draws you want, for example,

BOOTSTRAP = 100;

Francesca Pesola posted on Friday, May 04, 2012 - 10:05 am

Hi Linda,

Thanks for your promt response.

When I use:

Analysis:
bootstrap=100;

I still experience the same problem, i.e. the output does not produce chi square or CFI as model fit indices - but just the output as in my earlier message.

Is it because I am using WLSMV?

Linda K. Muthen posted on Friday, May 04, 2012 - 10:08 am

Please send the output and your license number to support@statmodel.com.

Stefanie Morgenroth posted on Friday, November 09, 2012 - 9:06 am

Hi,

I have the same problem.

I am using SEM to explore a mediation model. The observed variables are binary/ordinal .

I am using the default estimator: WLSMV.

I would like to use bootstrap on the data; but when I add analysis:
estimator=wlsmv;
bootstrap = 1000; , MPlus does not produce the model estimates. Well below is what I obtain from the output:

THE MODEL ESTIMATION TERMINATED NORMALLY

MODEL FIT INFORMATION

Number of Free Parameters 269

WRMR (Weighted Root Mean Square Residual)

Value 0.721

Can anyone tell me why I am receiving such input and whether it is possible to use the bootstrap option and still obtain the chi-square and CFI model fit indices?

Many thanks

Linda K. Muthen posted on Friday, November 09, 2012 - 11:29 am

Only the standard errors are bootstrapped not the fit indices. If you want the regular, not bootstrapped, fit statistics, run the analysis without the BOOTSTRAP option.

Stefanie Morgenroth posted on Thursday, September 26, 2013 - 2:16 am

Hey,

I have the same problem. Is it possible to run the analysis without the bootstrap option even the sample is very small (n=80). Or what do you mean with the sentence 'run the analysis without the BOOTSTRAP option'. But I thought, that I have to do bootstrapping, if my sample is to small?! So I am not sure what I have to report. Or is the information about the wrmr enough, if I use bootstrapping?

Thank you very much!!!

Linda K. Muthen posted on Thursday, September 26, 2013 - 12:01 pm

The BOOTSTRAP option applies to standard errors not fit statistics. If you want fit statistics, remove the BOOTSTRAP option and you will obtain non-bootstrapped fit statistics. Use the results from the bootstrapped standard errors along with the non-bootstrapped fit statistics. Your small sample will result in lower power.

Stefan P. posted on Monday, August 31, 2020 - 9:18 am

Hello Dres. M&M,

to make it clear for me: if I use BOOTSTRAP, I get neither CFI nor TLI, right?

Best regards
Stefan

Tihomir Asparouhov posted on Monday, August 31, 2020 - 5:11 pm

I don't think that's correct. Send your example to support@statmodel.com