Bootstrapping the EFA PreviousNext
Mplus Discussion > Exploratory Factor Analysis >
Message/Author
 Dio Itsara posted on Thursday, September 16, 2010 - 7:41 pm
Hi Dr. Linda

I'm interesting to conduct the CI and SE from bootstrapping applied to the EFA.

Can we use Mplus to conduct bootstrap exploratory factor analysis?


Many thanks
Dio
 Linda K. Muthen posted on Friday, September 17, 2010 - 8:04 am
The BOOTSTRAP option is not available with TYPE=EFA.
 Jorge Mahecha posted on Friday, October 05, 2018 - 1:19 pm
As several years have passed after this question was responded, I wanted to check again if this functionality is now available in Mplus 8. It seems like a useful thing to have when working with a very large sample (I have over 100.000 people, for example): drawing random samples of a given size (say 5% or 10%), and having then averages of item loadings so on the different factors.

A) Does this look a sensible approach for EFA to you, and why or why not?

B) I'm working on DIF for this sample and I'm doing MIMIC. Modification Indices are huge vary widely between samples. However, there is a stable group of questions that comes as "DIF suspect" between the different samples. Would you agree that something like bootstraping would also be useful in this scenario? Why or why not?

Thanks for your response.
 Jorge Mahecha posted on Friday, October 05, 2018 - 1:20 pm
As several years have passed after this question was responded, I wanted to check again if this functionality is now available in Mplus 8. It seems like a useful thing to have when working with a very large sample (I have over 100.000 people, for example): drawing random samples of a given size (say 5% or 10%), and having then averages of item loadings on the different factors.

A) Does this look a sensible approach for EFA to you, and why or why not?

B) I'm working on DIF for this sample and I'm doing MIMIC. Modification Indices are huge vary widely between samples. However, there is a stable group of questions that comes as "DIF suspect" between the different samples. Would you agree that something like bootstraping would also be useful in this scenario? Why or why not?

Thanks for your response.
 Tihomir Asparouhov posted on Friday, October 05, 2018 - 2:37 pm
EFA is not available yet with bootstrap. External montecarlo using user's guide setup on page 46 (ESEM) is available and could in principle be used to analyze multiple samples. Bootstrap in general uses a re-sample that is of the same size as the original sample. It does not use a smaller subsample. I would recommend using the entire sample with N=100000 for your EFA analysis.
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: