Influence diagnostics in a general class of beta regression models

In this article, we consider the issue of assessing influence of observations in the general class of beta regression models introduced by Simas et al. (Comput. Stat. Data Anal. 54:348-366, 2010), which is very useful in situations in which the response is restricted to the standard unit interval (0,1). Our results generalize those in Espinheira et al. (Comput. Stat. Data Anal. 52:4417-4431, 2008a; J. Appl. Stat. 35:407-419, 2008b), which only apply to linear beta regression models. We define some residuals, and a Portmanteau test for serial correlation. Further, some influence methods, such as the global, local, and total local influence of an individual and generalized leverage, are discussed. Moreover, we also derive the normal curvatures of local influence under various perturbation schemes. Finally, simulation results and an application to real data show the usefulness of our results.