Skip to content
Implements a numerically stable method of estimating logistic regressions which can be useful when convergence issues are likely or when a system has to be robust to such problems arising unexpectedly. It also works when there is endogeneity as well. The estimator is based on a contraction mapping result (see included PDF for details).
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
README.md
rlr_letter_zf.pdf
tslslogit.r

README.md

tslslogit

Implements a numerically stable method of estimating logistic regressions which can be useful when convergence issues are likely or when a system has to be robust to such problems arising unexpectedly. It also works when there is endogeneity as well. The estimator is based on a contraction mapping result (see included PDF, rlr_letter_zf.pdf, for details).

Currently, an R implementation is included in the repository, and I plan to add a Stata version as well.

R Readme

Includes a function, tslslogit. It takes as an argument "y" (a binary outcome), "r" (a vector of covariates), and "w" (a vector of instruments, which can be omitted if there is no endogeneity). It includes one option (noconstant) which determines whether to add a constant to the instrument and covariate vectors (by default, it does). The function returns the estimated parameter vector.

You can’t perform that action at this time.