Flexible Tweedie regression models for continuous data

TL;DR

This paper introduces quasi- and pseudo-likelihood methods for fitting Tweedie regression models, enhancing computational efficiency and flexibility in modeling continuous, skewed, and zero-inflated data.

Contribution

It proposes novel estimation approaches that simplify computation, extend the model class, and remove restrictions on the power parameter, improving flexibility and efficiency.

Findings

Quasi-likelihood method is asymptotically efficient.

Methods are computationally faster and easier to implement.

Estimates and coverage rates are comparable to maximum likelihood.

Abstract

Tweedie regression models provide a flexible family of distributions to deal with non-negative highly right-skewed data as well as symmetric and heavy tailed data and can handle continuous data with probability mass at zero. The estimation and inference of Tweedie regression models based on the maximum likelihood method are challenged by the presence of an infinity sum in the probability function and non-trivial restrictions on the power parameter space. In this paper, we propose two approaches for fitting Tweedie regression models, namely, quasi- and pseudo-likelihood. We discuss the asymptotic properties of the two approaches and perform simulation studies to compare our methods with the maximum likelihood method. In particular, we show that the quasi-likelihood method provides asymptotically efficient estimation for regression parameters. The computational implementation of the alternative methods is faster and easier than the orthodox maximum likelihood, relying on a simple Newton scoring algorithm. Simulation studies showed that the quasi- and pseudo-likelihood approaches present estimates, standard errors and coverage rates similar to the maximum likelihood method. Furthermore, the second-moment assumptions required by the quasi- and pseudo-likelihood methods enables us to extend the Tweedie regression models to the class of quasi-Tweedie regression models in the Wedderburn's style. Moreover, it allows to eliminate the non-trivial restriction on the power parameter space, and thus provides a flexible regression model to deal with continuous data. We provide \texttt{R} implementation and illustrate the application of Tweedie regression models using three data sets.