Application of Thin-plate Spline and Distributed Lag Non-linear Model to Describe the Interactive Effect of Two Predictors on Count Outcomes

This study illustrates a new version of the distributed lag non-linear model by integrating it with the thin-plate smoothing spline. The new model is called “the thin-plate distributed lag non-linear model” and helps explore the lagged interactive effect of two predictors on the outcome. Through simulation, it is found that the thin-plate distributed lag non-linear model has superior performance. We apply the proposed model to analyze the mutual effect of diurnal temperature and absolute humidity on daily mortality of realistic data.