Inference with joint models under misspecified random effects distributions

Authors

  • Abdus Sattar Department of Population and Quantitative Health Sciences Case Western Reserve University, 10900 Euclid Ave, Cleveland, OH, USA 44106
  • Sanjoy K Sinha School of Mathematics and Statistics, Carleton University, Ottawa, ON K1S 5B6 Canada

Keywords:

Frailty; Joint model; Longitudinal data; Misspecified random effect; Mixed model; Skew-normal distribution.

Abstract

Joint models are often used to analyze survival data with longitudinal covariates or biomarkers. Latent random effects that are used to describe the relationship between longitudinal and survival outcomes are typically assumed to follow a multivariate Gaussian distribution. A joint likelihood analysis of the data provides valid inferences under a correctly specified random effects distribution. However, the maximum likelihood method may produce biased estimators under a misspecified random effects distribution, and hence may provide invalid inferences. In this paper, we explore the empirical properties of the maximum likelihood estimators under various types of random effects, and propose a skewnormal distribution to address uncertainties in random effects. An extensive Monte Carlo study shows that our proposed method provides robust and efficient estimators under various types of model misspecifications. We also present an application of the proposed method using a large clinical dataset obtained from the genetic and inflammatory markers of sepsis (GenIMS) study.

Journal of Statistical Research 2021, Vol. 55, No. 1, pp. 187-205

Abstract
34
PDF
15

Downloads

Published

2021-12-09

How to Cite

Sattar, A. ., & Sinha, S. K. (2021). Inference with joint models under misspecified random effects distributions. Journal of Statistical Research, 55(1), 187–205. Retrieved from https://banglajol.info/index.php/JStR/article/view/56582

Issue

Section

Articles