Abstract:
PanJen provides users the opportunity to explore the relationship between a dependent variable and its covariates with minimal restrictions. The package offers an easy and data-driven way to choose a functional form in multiple linear regression models by comparing a range of parametric transformations. The parametric functional forms are benchmarked with a non-parametric smoothed relationship. The package allows users to generate plots that show the functional form relationship between the explanatory variable and the dependent variable. Furthermore, PanJen allows users to specify specific functional transformations, driven by an a priori and theory-based hypothesis. The plots and model fit metrics enable users to make an informed choice of how to specify the functional form the regression. We show that the PanJen ranking outperforms the Box-Tidwell transformation, especially in the presence of inefficiency, heteroscedasticity or endogeneity.