Biostatistics Journal Club: Estimating and Presenting Nonlinear and Interaction Effects with Restricted Cubic Splines

Most regression models commonly used in clinical and epidemiologic research–including logistic and Cox regression–rely on implicit assumptions of linearity and additivity. Restricted Cubic Splines (RCS) are a flexible tool that can greatly improve the model fit in the presence of non-linear associations. Including RCS transformations in regression models, however, presents challenges in terms of both implementation and results interpretation, especially when non-linear and non-additive (interaction) effects are both present.

In this talk, Andrea Bellavia, PhD, of Harvard Medical School and Harvard T.H. Chan School of Public Health, will introduce a flexible framework based on RCS to incorporate non-linear and interaction effects in common modeling approaches such as logistic and Cox regression. Extensive software material to facilitate the implementation and graphical presentation of RCS results in R, Stata, and SAS, will also be discussed and provided.

Required Reading:

Estimating and Presenting Hazard Ratios and Absolute Risks from a Cox Model with Complex Nonlinear Interactions

Speaker:

Andrea Bellavia, PhD