Description
glmwals computes the WALS estimates of univariate generalized linear models (GLMs) such as gamma, logit, probit, complementary log-log, and Poisson regressions.
The approach linearizes the likelihood equations, similar to the Newton–Raphson and Fisher scoring algorithms. Unlike the iteratively reweighted least squares (IRLS) procedure, which iterates until convergence, WALS is applied to the data transformations obtained from the first iteration.
A single iteration of the procedure yields the one-step WALS estimator. To reduce the dependence of this estimator on the starting values, glmwals by default performs an iterative procedure that repeatedly updates the starting values using the one-step WALS estimates from the previous iteration until a convergence criterion is satisfied.
Help files
After installation, you can view the estimation options by typing in Stata
help glmwals
and the post-estimation options by typing
help glmwals postestimation
Key references (in chronological order)
De Luca, G., Magnus, J. R., and Peracchi, F. (2018). Weighted-average least squares estimation of generalized linear models. Journal of Econometrics, 204, 1-17.
De Luca, G., and Magnus, J. R. (2025c). Weighted-average least squares: Beyond the simple linear regression model. The Stata Journal, 25(4), 1-40.