%0 Journal Article
%T Generalized two
%A Jesper N. Wulff
%J The Stata Journal
%@ 1536-8734
%D 2019
%R 10.1177/1536867X19854017
%X Researchers who model fractional dependent variables often need to consider whether their data were generated by a two-part process. Two-part models are ideal for modeling two-part processes because they allow us to model the participation and magnitude decisions separately. While community-contributed commands currently facilitate estimation of two-part models, no specialized command exists for fitting two-part models with process dependency. In this article, I describe generalized two-part fractional regression, which allows for dependency between modelsĄŻ parts. I show how this model can be fit using the community-contributed cmp command (Roodman, 2011, Stata Journal 11: 159¨C206). I use a data example on the financial leverage of firms to illustrate how cmp can be used to fit generalized two-part fractional regression. Furthermore, I show how to obtain predicted values of the fractional dependent variable and marginal effects that are useful for model interpretation. Finally, I show how to compute model fit statistics and perform the RESET test, which are useful for model evaluation
%K st0558
%K generalized two-part fractional regression
%K process dependence
%K fractional probit
%K cmp
%U https://journals.sagepub.com/doi/full/10.1177/1536867X19854017