Presiduals: An R package for residual analysis using probability-scale residuals

Abstract

We present the R package PResiduals for residual analysis using the probability-scale residual. This residual is well defined for a wide variety of outcome types and models, including some settings where other popular residuals are not applicable. It can be used for model diagnostics, tests of conditional associations, and covariate-adjustment for Spearman’s rank correlation. These tests and measures of conditional association are applicable to any orderable variable. They use order information but do not require assigning scores to ordered categorical variables or transforming continuous variables, and therefore, can achieve a good balance between robustness and efficiency. We illustrate the usage of the PResiduals package with a publicly available dataset.