95% CI for effect size (Est./S.E.)... PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Mario posted on Sunday, September 11, 2016 - 3:07 am
Dear users
we are working with SEM.
The output of the program give us the Estimate, S.E. and Est./S.E., which I interpret as to be the effect size.
But, how to obtain the 95% Confidence Intervals (CI) of the effect size of each predictor? when I use the output 'CInterval' I get the lower and upper 0.5%, 2.5% and 5% CI for the estimators. Should I divide these, for example, Lower 2.5% CI by the SE of the estimate?
For example, if Estimate=1.257; S.E.=0.373 and Effect size (Est./S.E.)= 3.365 (=1.257/0.373). And if Lower 2.5% CI of the estimate=0.525, can I say that Lower 2.5% CI of the effect size is 0.525/0.373??

thanks!
 Bengt O. Muthen posted on Sunday, September 11, 2016 - 10:45 am
Effect size is typically defined as a mean difference divided by their SD, not Est/SE.

So express that effect size in Model Constraint using model parameter labels and you will get a CI for it.

If you are referring to standardized effects (stand'd slopes) you find them and their CI's in the regular output.
 Mario posted on Monday, September 12, 2016 - 2:24 am
Thanks Dr. Muthen,
so, just to confirm my little knowledge. What you called standardized effects, is the Est./S.E provided by MPlus output, right? in Sweet and Grace-Martin (2011),I read that to report them "serves as standardized effect size statistics".

In the SAMPSTAT output I get Estimate, S.E., Est./S.E. & the two-tailed P-value. I f I use CINTERVAL I get the same plus the CONFIDENCE INTERVALS of the estimates, but to get the CI of the standardized effects, should I divide these values by the SE of the estimate (the same I get with SAMPSTAT)?

Thanks
 Bengt O. Muthen posted on Monday, September 12, 2016 - 9:17 am
Est/SE is not the standardized value. See the UG for a description of standardization.
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: