endogCopula: Gaussian Copula Based Endogeneity Corrections

The endogCopula package implements eight copula-based estimators for addressing endogeneity in linear models behind one interface. The six cross-sectional estimators are from Park and Gupta (2012) (CopRegPG()), Yang et al. (2025) (CopReg2sCOPE()), Hu et al. (2025) (CopReg2sCOPEnp()), Breitung et al. (2024) (CopRegBMW()), Haschka (2025) (CopRegIMA()), and Liengaard et al. (2025) (CopRegJAMS()); all six are reachable through the single entry point copreg().

Details

All cross-sectional estimators share the two-part formula interface y ~ endog1 + endog2 | exog1 + exog2, a cdf argument selecting one of seven marginal CDF estimators (kernel, empirical, and rank based), and bootstrap standard errors controlled via nboots. Fitted models are returned as "copreg" objects with print(), summary(), coef(), vcov(), residuals(), fitted(), predict(), update(), and confint() methods, and validity() reports skewness, kurtosis, and normality diagnostics for checking the estimators' assumptions.

The panel estimator and the Bayesian sampler of Haschka live in the companion packages endogCopulaPanel (CopRegPANEL()) and endogCopulaBayes (CopRegBAYES()); those packages call back into a handful of internal helpers exported from here (see ?endogCopula-internals) so that the model-parsing, CDF-estimation, and diagnostic code is not duplicated across the three packages.

See also

Useful links:

• https://ashgreat.github.io/endogCopula/

• https://github.com/ashgreat/endogCopula

• Report bugs at https://github.com/ashgreat/endogCopula/issues

Author

Maintainer: Ashwin Malshe ashwin@malshe.com

Other contributors:

• Rouven E. Haschka (Author of the reference implementation that the copula endogeneity estimators in this package are ported from) [contributor, copyright holder]