Errors in variables with clustering and fixed effects

[paulvonhippel]

We are wanting to fit a regression model that has both errors in variables and fixed effects.

If there were a command called xteivreg, we would use it, but there is no such command, right?

We see two options:
(1) Pre-process the data for fixed effects using xtreg, fe, then run eivreg
(2) Or run eivreg with a dummy variable for each fixed effect.
Is there another way?

Also...we would like to cluster our standard errors, but there is no cluster option in eivreg. Is there an alternative?

Many thanks,
Paul

[Rich Goldstein]

first, there is some controversy about -eivreg-; see Lockwood, JR, et al. (2017), " Errors-in-Variables Regression: Why Stata’s -eivreg- is Wrong and What To Do Instead ", SREE Spring Research Conference

Second, you might want to see the two part tutorial in the current issue of Statistics in Medicine: Keogh, RH, et al. (2020), "
STRATOS guidance document on measurement error and misclassification of variables in observational epidemiology: Part 1—Basic theory and simple methods of adjustment" Statistics in Medicine, 39: 2197-2231 and, Shaw, PA, et al., (2020), "
STRATOS guidance document on measurement error and misclassification of variables in observational epidemiology: Part 2—More complex methods of adjustment and advanced topics, Statistics in Medicine, 39: 2232-2263

[Stephen Jenkins]

Following up Rich's remark: a related Lockwood et al. paper has appeared in the Stata Journal:

Recommendations about estimating errors-in-variables regression in Stata

J. R. Lockwood, Daniel F. McCaffrey
First Published March 24, 2020 Research Article
https://doi.org/10.1177/1536867X20909692


Volume: 20 issue: 1, page(s): 116-130
Article first published online: March 24, 2020; Issue published: March 1, 2020

[paulvonhippel]

Thanks! Here's the abstract: Errors-in-variables (EIV) regression is a standard method for consistent estimation in linear models with error-prone covariates. The Stata commands eivreg and sem both can be used to compute the same EIV estimator of the regression coefficients. However, the commands do not use the same methods to estimate the standard errors of the estimated regression coefficients. In this article, we use analysis and simulation to demonstrate that standard errors reported by eivreg are negatively biased under assumptions typically made in latent-variable modeling, leading to confidence interval coverage that is below the nominal level. Thus, sem alone or eivreg augmented with bootstrapped standard errors should be preferred to eivreg alone in most practical applications of EIV regression.

[paulvonhippel]

So eivreg is OK with the bootstrap option. That answers my question about clustering.

What about fixed effects? My impression is that the only way to incorporate them is with dummy variables or by mean-centering using the xtdata, fe command.

It would be nice if there were an easier way, like a command called xteivreg, fe. Any such luck?

[Andreas Sihotang]

Hello, I am new to this forum. I am also trying to use eivreg with dummy for country and year fixed effects. I use eivreg command with dummy variables for both country and year, but STATA warns me that 'reliability r() too small'. Would the option suggested by Paul to use eivreg with a dummy variable for each fixed effect work?