ADF is WLS?

[fernando vazquez]

Dear friends of STATA,

I have version 13 of Stata and I wanted to ask you if the ADF method used by STATA for linear SEM is WLS? Or is DWLS like in Lisrel? Or is it WLSMV like in Mplus?

Thank you,


Fernando Vazquez

[Richard Williams]

I believe ADF in Stata sem is WLS in Mplus and WLS in R's lavaan. At least that is how I have coded in when switching between the programs.

[Richard Williams]

The Mplus 8 manual says "WLS – weighted least square parameter estimates with conventional standard errors and chi-square test statistic that use a full weight matrix. The WLS chi-square test statistic is also referred to as ADF when all outcome variables are continuous."

Also, "WLSMV – weighted least square parameter estimates using a diagonal weight matrix with standard errors and mean- and variance-adjusted chi-square test statistic that use a full weight matrix."

Mplus has all these zillions of options -- I don't know many are available in Stata. It gets confusing because different names get used for the same thing. Here is a section of code from my xtdpdml program that converts Stata options into Mplus options.

                * Next code checks for options that change MPlus estimation method
                if "`vce'" == "vce(robust)" {
                        local estimator MLR
                }
                else if "`vce'" == "vce(sbentler)" {
                        local estimator MLM
                }
                else if "`method'" == "method(adf)" {
                        local estimator WLS
                }
                else local estimator ML

For converting from Stata to lavaan I have

                ///Check for options that change lavaan standard errors
                if "`vce'" == "vce(robust)" {
                        local lavse robust
                }
                else local lavse default

                /// Check for method specified. 
                if "`method'" == "method(adf)" {
                        local lavestimator WLS
                }
                else local lavestimator ML
                
                /// Satorra Bentler gets separate treatment
                /// Satorra Bentler doesn't work the same in lavaan
                /// All you get is a test and robust standard errors
                if "`vce'" == "vce(sbentler)" {
                        local lavestimator MLM
                }

[fernando vazquez]

Thank you very much Richard
Why do you think Stata is not updated with a more effective ADF method?
You need a very large sample to use WLS .....

[Richard Williams]

Only so many hours in the day, I suppose. There are other enhancements to sem that would be higher priority for me personally, but you could put something on the Stata 17 wishlist if you want.

Mplus is ridiculously expensive, but if you desperately need the features only it offers it might be worth it.

[fernando vazquez]

Thank you very much Richard