Work

Examining the Effect of Non-Linearity on Missing Data in Observational Studies

Year: 2025

Type: article

Abstract: Dose-response curves in measure a range of chemical doses and responses. This paper demonstrates that choice of missing data techniques can affect results given non-linear relationships between exposures and outcomes. We created a dataset simulating a chemical exposure, personal characteristics, and chemical outcomes. The four outcomes included shapes common in dose-response studies (linear, polynomial, exponential, and change-point). We compared three different methods to account for the missing data: complete case (CC), substitution (S), and nested multiple imputation (NMI). Test scenarios included varying the level of noise and percentage of missing data then bias at different points in the curve was compared across different scenarios.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.15390425

Author: Michelle Mellers

Institutions: Henry M. Jackson Foundation, Uniformed Services University of the Health Sciences

Language: English

Topic: Statistical Methods and Bayesian Inference

Subfield: Statistics and Probability

Field: Mathematics

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W6949886999

FWCI:

Cites:

Cited by:

Related to: