Discrepancy between bootstrapped CI a... PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Greg Thrasher posted on Wednesday, July 03, 2019 - 3:06 pm
Hello,

I am running a mediation model using SEM. Using a bootstrap of 1000 my unstandardized a path shows a non-sig p value (p = .08). However the unstandardized CI for that path coefficient (m on x) does not contain 0. The unstandardized CI for the indirect effect also does not contain 0.

When I run without bootstrapping the a path is significant based on the p-value at p = .01.

How can I reconcile this discrepancy between the p-value and confidence interval?

I also have run this with estimator = ML and get some slightly different results. The bootstraps take a long time using ML however.

Thanks so much for your time.
 Bengt O. Muthen posted on Wednesday, July 03, 2019 - 3:42 pm
p-values are based on Est +- 1.96* SE, that is, they refer to symmetric CIs. Bootstrapping gives you non-symmetric CIs which is usually what you want.
 Greg Thrasher posted on Wednesday, July 03, 2019 - 3:46 pm
So in this case would you suggest using the CIs to infer significance as opposed to the p-value?
 Bengt O. Muthen posted on Wednesday, July 03, 2019 - 5:12 pm
Yes.
 Greg Thrasher posted on Wednesday, July 03, 2019 - 5:23 pm
Thank you for the quick and helpful response as always!
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: