Principled methods with which to appropriately analyze missing data have long existed; however, broad implementation of these methods remains challenging. In this and 2 companion papers (Am J Epidemiol. 2018;187(3):576-584 and Am J Epidemiol. 2018;187(3):585-591), we discuss issues pertaining to missing data in the epidemiologic literature. We provide details regarding missing-data mechanisms and nomenclature and encourage the conduct of principled analyses through a detailed comparison of multiple imputation and inverse probability weighting. Data from the Collaborative Perinatal Project, a multisite US study conducted from 1959 to 1974, are used to create a masked data-analytical challenge with missing data induced by known mechanisms. We illustrate the deleterious effects of missing data with naive methods and show how principled methods can sometimes mitigate such effects. For example, when data were missing at random, naive methods showed a spurious protective effect of smoking on the risk of spontaneous abortion (odds ratio (OR) = 0.43, 95% confidence interval (CI): 0.19, 0.93), while implementation of principled methods multiple imputation (OR = 1.30, 95% CI: 0.95, 1.77) or augmented inverse probability weighting (OR = 1.40, 95% CI: 1.00, 1.97) provided estimates closer to the "true" full-data effect (OR = 1.31, 95% CI: 1.05, 1.64). We call for greater acknowledgement of and attention to missing data and for the broad use of principled missing-data methods in epidemiologic research.

Epidemiologic studies are frequently susceptible to missing information. Omitting observations with missing variables remains a common strategy in epidemiologic studies, yet this simple approach can often severely bias parameter estimates of interest if the values are not missing completely at random. Even when missingness is completely random, complete-case analysis can reduce the efficiency of estimated parameters, because large amounts of available data are simply tossed out with the incomplete observations. Alternative methods for mitigating the influence of missing information, such as multiple imputation, are becoming an increasing popular strategy in order to retain all available information, reduce potential bias, and improve efficiency in parameter estimation. In this paper, we describe the theoretical underpinnings of multiple imputation, and we illustrate application of this method as part of a collaborative challenge to assess the performance of various techniques for dealing with missing data (Am J Epidemiol. 2018;187(3):568-575). We detail the steps necessary to perform multiple imputation on a subset of data from the Collaborative Perinatal Project (1959-1974), where the goal is to estimate the odds of spontaneous abortion associated with smoking during pregnancy.

The development of coherent missing data models to account for nonmonotone missing at random (MAR) data by inverse probability weighting (IPW) remains to date largely unresolved. As a consequence, IPW has essentially been restricted for use only in monotone missing data settings. We propose a class of models for nonmonotone missing data mechanisms that spans the MAR model, while allowing the underlying full data law to remain unrestricted. For parametric specifications within the proposed class, we introduce an unconstrained maximum likelihood estimator for estimating the missing data probabilities which can be easily implemented using existing software. To circumvent potential convergence issues with this procedure, we also introduce a Bayesian constrained approach to estimate the missing data process which is guaranteed to yield inferences that respect all model restrictions. The efficiency of the standard IPW estimator is improved by incorporating information from incomplete cases through an augmented estimating equation which is optimal within a large class of estimating equations. We investigate the finite-sample properties of the proposed estimators in a simulation study and illustrate the new methodology in an application evaluating key correlates of preterm delivery for infants born to HIV infected mothers in Botswana, Africa.

Knowledge of biological relatedness between samples is important for many genetic studies. In large-scale human genetic association studies, the estimated kinship is used to remove cryptic relatedness, control for family structure, and estimate trait heritability. However, estimation of kinship is challenging for sparse sequencing data, such as those from off-target regions in target sequencing studies, where genotypes are largely uncertain or missing. Existing methods often assume accurate genotypes at a large number of markers across the genome. We show that these methods, without accounting for the genotype uncertainty in sparse sequencing data, can yield a strong downward bias in kinship estimation. We develop a computationally efficient method called SEEKIN to estimate kinship for both homogeneous samples and heterogeneous samples with population structure and admixture. Our method models genotype uncertainty and leverages linkage disequilibrium through imputation. We test SEEKIN on a whole exome sequencing dataset (WES) of Singapore Chinese and Malays, which involves substantial population structure and admixture. We show that SEEKIN can accurately estimate kinship coefficient and classify genetic relatedness using off-target sequencing data down sampled to ~0.15X depth. In application to the full WES dataset without down sampling, SEEKIN also outperforms existing methods by properly analyzing shallow off-target data (~0.75X). Using both simulated and real phenotypes, we further illustrate how our method improves estimation of trait heritability for WES studies.

No abstract