Confidence intervals for StdYX PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Jen Nickelson posted on Saturday, September 15, 2007 - 7:29 pm
Is there some way to calculate CIs for StdYX or another way to determine if the standardized regression coefficient is statistically significant?
 Linda K. Muthen posted on Tuesday, September 18, 2007 - 10:31 am
You can use MODEL CONSTRAINT to define the standardized parameter and you will obtain a standard error for the parameter.
 Jen Nickelson posted on Friday, September 21, 2007 - 5:54 pm
Thank you for your reply. I think I was able to write the program to name the parameters, but I'm afraid I can not figure out how to write the model constraint command to get the standard errors for the parameters.
 Linda K. Muthen posted on Friday, September 21, 2007 - 9:40 pm
If you specify the parameters in MODEL CONSTRAINT, you will automatically obtain the standard errors.
 Jen Nickelson posted on Friday, September 28, 2007 - 11:57 am
Forgive my ignorance . . . I'm a dietitian not a statistician. Looking at example 5.20 in the manual, the formula for a standardized estimate appears to be:

standardized estimate = factor loading *square root of factor variance divided by the square root of (factor loading **2*factor variance)

What would the formula be for the standardized estimate for regression coefficients?
 Linda K. Muthen posted on Friday, September 28, 2007 - 2:20 pm
To standardize a regression coefficient with an observed dependent variable and an observed independent variable, multiply the regression coefficient by the standard deviation of x and divide it by the model estimated standard deviation of y.

Note that the formula you give above is not correct. See Example 5.20 for the full formula.
 MAH posted on Wednesday, October 01, 2008 - 8:24 pm
I am trying to obtain CI's for a twin analysis of a binary variable. The unstandardized and standarized estimates for A,C, and E are exactly the same and so are the CI's for the standardized and unstandarized estimates. The CI's go below 0 and above 1, suggesting that, for example, genes account for negative proportions of variance or more than 100% of the variance in the trait. This is quite fishy. Can you explain?
 Linda K. Muthen posted on Thursday, October 02, 2008 - 3:56 pm
In a twin model, you fix the factor variances to one and free the factor loading. This results in working with correlations. This is why your raw and standardized results are the same.

The size of the confidence interval is determined by sample size and the size of the standard error. If you are not using Version 5.1, you should download it.
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: