Copula based generalized additive models with non-random sample selection

TL;DR

This paper introduces a novel copula-based generalized additive model that effectively addresses non-random sample selection, allowing flexible distribution choices, dependence structures, and nonparametric effects, with proven asymptotic properties and validated through simulations.

Contribution

It extends generalized additive models to account for non-random sample selection using copulae and penalized likelihood, providing a flexible and theoretically grounded approach.

Findings

The method accurately models non-random sample selection.

Asymptotic theory for the estimators is established.

Simulation studies demonstrate the approach's effectiveness.

Abstract

Non-random sample selection is a commonplace amongst many empirical studies and it appears when an output variable of interest is available only for a restricted non-random sub-sample of data. We introduce an extension of the generalized additive model which accounts for non-random sample selection by using a selection equation. The proposed approach allows for different distributions of the outcome variable, various dependence structures between the (outcome and selection) equations through the use of copulae, and nonparametric effects on the responses. Parameter estimation is carried out within a penalized likelihood and simultaneous equation framework. We establish asymptotic theory for the proposed penalized spline estimators, which extends the recent theoretical results for penalized splines in generalized additive models, such as those by Kauermann et al. (2009) and Yoshida & Naito (2014). The empirical effectiveness of the approach is demonstrated through a simulation study.