Bootstrapped standard errors PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Daniel Rodriguez posted on Tuesday, June 26, 2012 - 11:33 am
Hello,
I recently conducted an analysis in which I used bootstrapped standard errors to test for mediation. Prior to bootstrapping, the direct effect of my predictor variable to the outcome variable was significant, as was the indirect effect of interest. After bootstrapping with 1,000 repetitions, the indirect effect remained significant however the direct effect was no longer significant. So, which direct effect is correct, the direct effect prior to bootstrapping or after bootstrapping? Why is there a change to the direct effect?
 Bengt O. Muthen posted on Tuesday, June 26, 2012 - 11:44 am
Which estimator did you use in the non-bootstrap run?
 Daniel Rodriguez posted on Tuesday, June 26, 2012 - 11:51 am
WLSMV
 Daniel Rodriguez posted on Wednesday, June 27, 2012 - 4:13 am
Is that OK?
 Bengt O. Muthen posted on Wednesday, June 27, 2012 - 10:25 am
Bootstrap SEs are often a little bigger, which can explain the difference with regular WLSMV. You can also use Bayes to see how that comes out.
 Daniel Rodriguez posted on Wednesday, June 27, 2012 - 11:15 am
Thanks
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: