Monotonic Nonparametric Dose Response Model

TL;DR

This paper introduces a Bayesian nonparametric model using Alamri Monotonic splines for dose-response analysis, ensuring monotonicity and positivity, demonstrated on simulated and real pesticide data.

Contribution

It presents a novel Bayesian nonparametric approach with AM-splines for monotonic dose-response modeling, addressing limitations of parametric methods.

Findings

Effective modeling of monotonic dose-response relationships.

Successful application to pesticide research data.

Flexible nonparametric approach with Bayesian inference.

Abstract

Toxicologists are often concerned with determining the dosage to which an individual can be exposed with an acceptable risk of adverse effect. These types of studies have been conducted widely in the past, and many novel approaches have been developed. Parametric techniques utilizing ANOVA and nonlinear regression models are well represented in the literature. The biggest drawback of parametric approaches is the need to specify the correct model. Recently, there has been an interest in nonparametric approaches to tolerable dosage estimation. In this work, we focus on the monotonically decreasing dose response model where the response is a percent to control. This poses two constraints to the nonparametric approach. The doseresponse function must be one at control (dose = 0), and the function must always be positive. Here we propose a Bayesian solution to this problem using a novel class of nonparametric models. A basis function developed in this research is the Alamri Monotonic spline (AM-spline). Our approach is illustrated using both simulated data and an experimental dataset from pesticide related research at the US Environmental Protection Agency.