Using Bayes function of Stata 14 to deal with measurement error problem

[Lei Lu]

I am running a linear regression model for which some of independent variables have measurement error problem. Someone suggested me to use Bayesian MCMC and fortunately Stata 14 has this function. Who knows how to use the Bayes function to deal with the measurement error? Thanks...

[Joseph Coveney]

For measurement error in the predictors, you might want to look into structural equation modeling (SEM). If you Google around (e.g., stata AND sem AND "measurement error"), you can find what look like tutorials in using Stata's sem for accommodating measurement error in the predictors, for example, here, although I haven't taken a look at it. The user's manual entry for sem has numerous examples, too. Perhaps a few of them deal specifically with your problem.

[wbuchanan]

As a fairly late aside, John Lockwood and a few of his colleagues from RAND/ETS did a talk at SREE a couple years ago about error-in-variables models. It seems that the big factor is whether the value that can be supplied are the error variances or reliability coefficients: https://www.sree.org/conferences/201...pdf&item=slide