Multiple Imputation for Causal Inference
نویسندگان
چکیده
The potential outcome framework for causal inference is fundamentally a missing data problem with a special, the so-called file-matching, pattern of missing data. Given the large body of literature on various methods for handling missing data and associated software, it will be useful to use such methods to facilitate causal inference for routine applications. This article uses the sequential regression or chained equation imputation methodology for handling missing data to impute the potential outcomes based on the observed data. The causal inference parameters are formulated based on the models for the completed data and standard multiple imputation combining rules are applied to infer about the direct and mediated effects. Since the special pattern of missing data makes certain parameters of the joint distribution not estimable, the multiple imputation framework is modified to incorporate constraints or prior information in terms of augmented complete-data. Given the ability of the multiple imputation framework to to handle several types of variables, missing values in covariates and the availability of software for performing multiple imputations, this approach makes easier to perform causal inference from both observational and randomized studies. The methodology is illustrated through an application aimed to understand and quantify direct and mediated effect of diabetes on the cardiovascular disease using the NHANES data. keywords: Direct Effect, Indirect Effect, Mediation, Observational Studies, Potential Outcomes, Randomized Studies
منابع مشابه
Causal Inference for Multilevel Networks of Early Ethnographically Well - studied Populations
Ethnographically well-studied cases have long been the concern and targets of attempts at causal inference that recognize network interdependencies (e.g., Dow, White and Burton 1982) – given that societies are spatially and historically interdependent. This paper uses methods for estimating Instrumental Variables (IV) for network effects of shared historical-linguistic and spatial proximities (...
متن کاملOptimal Causal Imputation for Control
The widespread applicability of analytics in cyberphysical systems has motivated research into causal inference methods. Predictive estimators are not sufficient when analytics are used for decision making; rather, the flow of causal effects must be determined. Generally speaking, these methods focus on estimation of a causal structure from experimental data. In this paper, we consider the dual...
متن کاملPenalized regression procedures for variable selection in the potential outcomes framework.
A recent topic of much interest in causal inference is model selection. In this article, we describe a framework in which to consider penalized regression approaches to variable selection for causal effects. The framework leads to a simple 'impute, then select' class of procedures that is agnostic to the type of imputation algorithm as well as penalized regression used. It also clarifies how mo...
متن کاملBayesian Methods to Impute Missing Covariates for Causal Inference and Model Selection
BAYESIAN METHODS TO IMPUTE MISSING COVARIATES FOR CAUSAL INFERENCE AND MODEL SELECTION by Robin Mitra Department of Statistical Science Duke University
متن کاملMultiple imputation methods for treatment noncompliance and nonresponse in randomized clinical trials.
Randomized clinical trials are a powerful tool for investigating causal treatment effects, but in human trials there are oftentimes problems of noncompliance which standard analyses, such as the intention-to-treat or as-treated analysis, either ignore or incorporate in such a way that the resulting estimand is no longer a causal effect. One alternative to these analyses is the complier average ...
متن کامل