Marginal analysis of longitudinal ordinal data with misclassification in both response and covariates

Abstract

Marginal methods have been widely used for the analysis of longitudinal ordinal and categorical data. These models do not require full parametric assumptions on the joint distribution of repeated response measurements but only specify the marginal or even association structures. However, inference results obtained from these methods often incur serious bias when variables are subject to error. In this paper, we tackle the problem that misclassification exists in both response and categorical covariate variables. We develop a marginal method for misclassification adjustment, which utilizes second-order estimating functions and a functional modeling approach, and can yield consistent estimates and valid inference for mean and association parameters. We propose a two-stage estimation approach for cases in which validation data are available. Our simulation studies show good performance of the proposed method under a variety of settings. Although the proposed method is phrased to data with a longitudinal design, it also applies to correlated data arising from clustered and family studies, in which association parameters may be of scientific interest. The proposed method is applied to analyze a dataset from the Framingham Heart Study as an illustration.

Correlated and misclassified binary observations in complex surveys

“…Misclassifications in binary covariates from complex surveys are another related problem often encountered in practice. This has been well documented by Ogburn & VanderWeele (2012), Chen, Yi & Wu (2014) and Yi (2017), among others. Our proposed estimation strategy does not apply directly to model parameters involving misclassified binary covariates for survey data.…”

“…Most of the articles, however, consider the non-survey setting. Chen, Yi & Wu (2011) developed a GEE approach to handle correlated binary response with misclassification; Chen, Yi & Wu (2014) further extended the method to deal with misclassified covariates; Meyer & Mittag (2017) summarized the recent misclassification models for binary choices in econometrics. This is the motivation behind our research to extend the GEE method proposed by Chen, Yi & Wu (2011) to handle correlated and misclassified binary observations for complex surveys (COMBOS) using the pseudo-GEE approach of Carrillo-Garcia, Chen & Wu (2010).…”

On The Interplay between Exposure Misclassification and Informative Cluster Size

“…Exposure misclassification and measurement error in clustered‐correlated settings have a rich history in the statistical literature (Carroll et al ., 2006; Yi, 2016). Numerous extensions to generalized estimating equations (GEEs) have been proposed to permit marginal inference in the presence of mismeasured covariates, including expected estimating equations (EEEs) (Wang and Pepe, 2000; Wang et al ., 2008), corrected estimating equations (Yi, 2005; Yi et al ., 2012; Chen et al ., 2014), regression calibration (Sánchez et al ., 2009) and simulation–extrapolation (Yi, 2008). Methods for GLMMs have similarly been proposed (Carroll et al ., 1997; Wang et al ., 1998; Lin and Carroll, 1999; Wang et al ., 1999; Liang, 2009; Yi et al ., 2011).…”

Instrumental variable estimation in ordinal probit models with mismeasured predictors

“…For example, see Huang & Wang (), Xu, Ma & Wang () and Yi (), and the references therein. Some authors, such as Li & Hsiao (), Abarin & Wang () and Li & Wang (), have studied this problem in the context of generalized linear models, while Chen, Yi & Wu () investigated this misclassification problem in a longitudinal data model with both a categorical response and covariates.…”