Double Robustness Research Articles

Semi-Supervised Triply Robust Inductive Transfer Learning

In this work, we propose a Semi-supervised Triply Robust Inductive transFer LEarning (STRIFLE) approach, which integrates heterogeneous data from a label-rich source population and a label-scarce target population and uses a large amount of unlabeled data simultaneously to improve the learning accuracy in the target population. Specifically, we consider a high dimensional covariate shift setting and employ two nuisance models, a density ratio model and an imputation model, to combine transfer learning and surrogate-assisted semi-supervised learning strategies effectively and achieve triple robustness. While the STRIFLE approach assumes the target and source populations to share the same conditional distribution of outcome Y given both the surrogate features S and predictors X , it allows the true underlying model of Y⏧ X to differ between the two populations due to the potential covariate shift in S and X . Different from double robustness, even if both nuisance models are misspecified or the distribution of Y⏧ S , X is not the same between the two populations when the shifted source population and the target population share enough similarities, the triply robust STRIFLE estimator can still partially use the source population when the shifted source population and the target population share enough similarities. Moreover, it is guaranteed to be no worse than the target-only surrogate-assisted semi-supervised estimator with an additional error term from transferability detection. These desirable properties of our estimator are established theoretically and verified in finite samples via extensive simulation studies. We use the STRIFLE estimator to train a Type II diabetes polygenic risk prediction model for the African American target population by transferring knowledge from electronic health records linked genomic data observed in a larger European source population. Supplementary materials for this article are available online, including a standardized description of the materials available for reproducing the work.

Spatial heterogeneity analysis of biased land resource supply policies on housing prices and innovation efficiency

To rationally allocate land resources, the government has formulated biased land resource supply policies. However, this policy will promote the fluctuation of regional housing prices, thereby affecting regional innovation efficiency. To elucidate the intrinsic logical relationship among these factors, this study investigates the influence of biased land resource supply policies on housing prices and innovation across 31 Chinese provinces over a 16-year period. It places particular emphasis on probing into the spatial disparities in this impact. The innovation of this paper lies in the development of a theoretical analysis model, which creatively utilizes the regional land supply status as an instrumental variable to address the endogeneity problem. At the same time, a double robustness test was conducted by substituting the explanatory variables and employing the fixed effects model. The results of the TSLS regression indicate: (1) The land resource supply policy, which prioritizes the central and western regions, has resulted in an increase in housing prices across the majority of regions in China, with the eastern region experiencing a more rapid rise compared to the central and western regions. (2) The accelerated growth of housing prices in the eastern region is likely to exert a more pronounced inhibitory effect on innovation compared to the central and western regions. In summary, China’s land resource supply policy, which favors the central and western regions, has exacerbated the rise in housing prices in the eastern region, and rising housing prices will inhibit the innovation efficiency of the region. Hence, it is imperative for the government to continuously optimize and adjust land resource supply policies, ensuring their alignment with population migration patterns, stabilizing the real estate market, while also considering the innovation efficiency of regional development.

Propensity weighting plus adjustment in proportional hazards model is not doubly robust.

Recently, it has become common for applied works to combine commonly used survival analysis modeling methods, such as the multivariable Cox model and propensity score weighting, with the intention of forming a doubly robust estimator of an exposure effect hazard ratio that is unbiased in large samples when either the Cox model or the propensity score model is correctly specified. This combination does not, in general, produce a doubly robust estimator, even after regression standardization, when there is truly a causal effect. We demonstrate via simulation this lack of double robustness for the semiparametric Cox model, the Weibull proportional hazards model, and a simple proportional hazards flexible parametric model, with both the latter models fit via maximum likelihood. We provide a novel proof that the combination of propensity score weighting and a proportional hazards survival model, fit either via full or partial likelihood, is consistent under the null of no causal effect of the exposure on the outcome under particular censoring mechanisms if either the propensity score or the outcome model is correctly specified and contains all confounders. Given our results suggesting that double robustness only exists under the null, we outline 2 simple alternative estimators that are doubly robust for the survival difference at a given time point (in the above sense), provided the censoring mechanism can be correctly modeled, and one doubly robust method of estimation for the full survival curve. We provide R code to use these estimators for estimation and inference in the supporting information.

Distribution-Free Prediction Intervals Under Covariate Shift, With an Application to Causal Inference

Owing to its appealing distribution-free feature, conformal inference has become a popular tool for constructing prediction intervals with a desired coverage rate. In scenarios involving covariate shift, where the shift function needs to be estimated from data, many existing methods resort to data-splitting techniques. However, these approaches often lead to wider intervals and less reliable coverage rates, especially when dealing with finite sample sizes. To address these challenges, we propose methods based on a pivotal quantity derived under a parametric working model and employ a resampling-based framework to approximate its distribution. The resampling-based approach can produce prediction intervals with a desired coverage rate without splitting the data and can be easily applied to causal inference settings where a shift in the covariate distribution can occur between treatment and control arms. Additionally, the proposed approaches enjoy a double robustness property and are adaptable to different prediction tasks. Our extensive numerical experiments demonstrate that, compared to existing methods, the proposed novel approaches can produce substantially shorter conformal prediction intervals with lower variability in the interval lengths while maintaining promising coverage rates and advantages in versatile usage. Supplementary materials for this article are available online, including a standardized description of the materials available for reproducing the work.

Weighting estimation in the cause-specific Cox regression with partially missing causes of failure.

Complex diseases are often analyzed using disease subtypes classified by multiple biomarkers to study pathogenic heterogeneity. In such molecular pathological epidemiology research, we consider a weighted Cox proportional hazard model to evaluate the effect of exposures on various disease subtypes under competing-risk settings in the presence of partially or completely missing biomarkers. The asymptotic properties of the inverse and augmented inverse probability-weighted estimating equation methods are studied with a general pattern of missing data. Simulation studies have been conducted to demonstrate the double robustness of the estimators. For illustration, we applied this method to examine the association between pack-years of smoking before the age of 30 and the incidence of colorectal cancer subtypes defined by a combination of four tumor molecular biomarkers (statuses of microsatellite instability, CpG island methylator phenotype, BRAF mutation, and KRAS mutation) in the Nurses' Health Study cohort.

Nonparametric estimation of conditional incremental effects

Abstract Conditional effect estimation has great scientific and policy importance because interventions may impact subjects differently depending on their characteristics. Most research has focused on estimating the conditional average treatment effect (CATE). However, identification of the CATE requires that all subjects have a non-zero probability of receiving treatment, or positivity, which may be unrealistic in practice. Instead, we propose conditional effects based on incremental propensity score interventions, which are stochastic interventions where the odds of treatment are multiplied by some factor. These effects do not require positivity for identification and can be better suited for modeling scenarios in which people cannot be forced into treatment. We develop a projection approach and a flexible nonparametric estimator that can each estimate all the conditional effects we propose and derive model-agnostic error guarantees showing that both estimators satisfy a form of double robustness. Further, we propose a summary of treatment effect heterogeneity and a test for any effect heterogeneity based on the variance of a conditional derivative effect and derive a nonparametric estimator that also satisfies a form of double robustness. Finally, we demonstrate our estimators by analyzing the effect of intensive care unit admission on mortality using a dataset from the (SPOT)light study.

Estimating dynamic treatment regimes for ordinal outcomes with household interference: Application in household smoking cessation.

The focus of precision medicine is on decision support, often in the form of dynamic treatment regimes, which are sequences of decision rules. At each decision point, the decision rules determine the next treatment according to the patient's baseline characteristics, the information on treatments and responses accrued by that point, and the patient's current health status, including symptom severity and other measures. However, dynamic treatment regime estimation with ordinal outcomes is rarely studied, and rarer still in the context of interference - where one patient's treatment may affect another's outcome. In this paper, we introduce the weighted proportional odds model: a regression based, approximate doubly-robust approach to single-stage dynamic treatment regime estimation for ordinal outcomes. This method also accounts for the possibility of interference between individuals sharing a household through the use of covariate balancing weights derived from joint propensity scores. Examining different types of balancing weights, we verify the approximate double robustness of weighted proportional odds model with our adjusted weights via simulation studies. We further extend weighted proportional odds model to multi-stage dynamic treatment regime estimation with household interference, namely dynamic weighted proportional odds model. Lastly, we demonstrate our proposed methodology in the analysis of longitudinal survey data from the Population Assessment of Tobacco and Health study, which motivates this work. Furthermore, considering interference, we provide optimal treatment strategies for households to achieve smoking cessation of the pair in the household.

Doubly robust estimation of multivariate fractional outcome means with multivalued treatments

. This article suggests a doubly robust method of estimating potential outcome means for multivariate fractional outcomes when the treatment of interest is unconfounded and can take more than two values. The method involves maximizing a propensity score weighted multinomial quasi-log-likelihood function with a multinomial logit conditional mean. We show that this estimator, which we call weighted multivariate fractional logit (wmflogit), consistently estimates the potential outcome means if either the propensity score model or the conditional mean model is misspecified. Our simulations demonstrate this double robustness property for the case of shares generated using a Dirichlet distribution. Finally, we advocate for the use of wmflogit by applying it to estimate time-use shares of women participating in the Mexican conditional cash transfer program, Progresa, using Stata’s fmlogit command developed by Buis.

Doubly Flexible Estimation under Label Shift

In studies ranging from clinical medicine to policy research, complete data are usually available from a population P , but the quantity of interest is often sought for a related but different population Q which only has partial data. We consider the setting when both outcome Y and covariate X are available from P but only X is available from Q , under the label shift assumption; i.e., the conditional distribution of X given Y is the same in the two populations. To estimate the parameter of interest in Q by leveraging information from P , three ingredients are essential: (a) the common conditional distribution of X given Y, (b) the regression model of Y given X in P , and (c) the density ratio of the outcome Y between the two populations. We propose an estimation procedure that only needs some standard nonparametric technique to approximate the conditional expectations with respect to (a), while by no means needs an estimate or model for (b) or (c); i.e., doubly flexible to the model misspecifications of both (b) and (c). This is conceptually different from the well-known doubly robust estimation in that, double robustness allows at most one model to be misspecified whereas our proposal can allow both (b) and (c) to be misspecified. This is of particular interest in label shift because estimating (c) is difficult, if not impossible, by virtue of the absence of the Y-data from Q . While estimating (b) is occasionally off-the-shelf, it may encounter issues related to the curse of dimensionality or computational challenges. We develop the large sample theory for the proposed estimator, and examine its finite-sample performance through simulation studies as well as an application to the MIMIC-III database.

Estimating curvilinear time-varying treatment effects: Combining g-estimation of structural nested mean models with time-varying effect models for longitudinal causal inference.

Longitudinal designs can fortify causal inquiries of a focal predictor (i.e., treatment) on an outcome. But valid causal inferences are complicated by causal feedback between confounders and treatment over time. G-estimation of a structural nested mean model (SNMM) is designed to handle the complexities beset by measured time-varying or treatment-dependent confounding in longitudinal data. But valid inference requires correctly specifying the functional form of the SNMM, such as how the effects stay constant or change over time. In this article, we develop a g-estimation strategy for linear structural nested mean models whose causal parameters adopt the form of time-varying coefficient functions. These time-varying coefficient functions are smooth semiparametric functions of time that permit probing how the treatment effects may change curvilinearly. Further effect modification by time-invariant and time-varying covariates can be readily postulated in the SNMM to test fine-grained effect heterogeneity. We then describe a g-estimation strategy for estimating such an SNMM. We utilize the established time-varying effect model (TVEM) approach from the prevention and psychotherapy research literature for modeling flexible changes in covariate-outcome associations over time. Moreover, we exploit a known benefit of g-estimation over routine regression methods: its double robustness conferring protection against biases induced by certain forms of model misspecification. We encourage psychology researchers seeking correct causal conclusions from longitudinal data to use an SNMM with time-varying coefficient functions to assess curvilinear causal effects over time, and to use g-estimation with TVEM to resolve measured treatment-dependent confounding. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Outlier-Resistant Estimators for Average Treatment Effect in Causal Inference

The inverse probability (IPW) and doubly robust (DR) estimators are often used to estimate the average causal effect (ATE), but are vulnerable to outliers. The IPW/DR median can be used for outlier-resistant estimation of the ATE, but the outlier resistance of the median is limited and it is not resistant enough for heavy contamination. We propose extensions of the IPW/DR estimators with density power weighting, which can eliminate the influence of outliers almost completely. The outlier resistance of the proposed estimators is evaluated through the unbiasedness of the estimating equations. Unlike the median-based methods, our estimators are resistant to outliers even under heavy contamination. Interestingly, the naive extension of the DR estimator requires bias correction to keep the double robustness even under the most tractable form of contamination. In addition, the proposed estimators are found to be highly resistant to outliers in more difficult settings where the contamination ratio depends on the covariates. The outlier resistance of our estimators from the viewpoint of the influence function is also favorable. Our theoretical results are verified via Monte Carlo simulations and real data analysis. The proposed methods were found to have more outlier resistance than the median-based methods and estimated the potential mean with a smaller error than the median-based methods.

Double Robustness of Local Projections and Some Unpleasant Varithmetic

Double Robustness of Local Projections and Some Unpleasant Varithmetic

Efficient and Robust Estimation of the Generalized LATE Model

This article studies the estimation of causal parameters in the generalized local average treatment effect (GLATE) model, which expands upon the traditional LATE model to include multivalued treatments. We derive the efficient influence function (EIF) and the semiparametric efficiency bound (SPEB) for two types of causal parameters: the local average structural function (LASF) and the local average structural function for the treated (LASFT). The moment conditions generated by the EIF satisfy two robustness properties: double robustness and Neyman orthogonality. Based on the robust moment conditions, we propose the double/debiased machine learning (DML) estimator for estimating the LASF. The DML estimator is well-suited for high dimensional settings. We also propose null-restricted inference methods that are robust against weak identification issues. As an empirical application of these methods, we examine the potential health outcome across different types of health insurance plans using data from the Oregon Health Insurance Experiment.

Efficient evaluation of natural stochastic policies in off-line reinforcement learning

Summary We study the efficient off-policy evaluation of natural stochastic policies, which are defined in terms of deviations from the unknown behaviour policy. This is a departure from the literature on off-policy evaluation that largely considers the evaluation of explicitly specified policies. Crucially, off-line reinforcement learning with natural stochastic policies can help alleviate issues of weak overlap, lead to policies that build upon current practice and improve policies’ implementability in practice. Compared with the classic case of a prespecified evaluation policy, when evaluating natural stochastic policies, the efficiency bound, which measures the best-achievable estimation error, is inflated since the evaluation policy itself is unknown. In this paper we derive the efficiency bounds of two major types of natural stochastic policies: tilting policies and modified treatment policies. We then propose efficient nonparametric estimators that attain the efficiency bounds under lax conditions and enjoy a partial double robustness property.

Selective machine learning of doubly robust functionals

Summary While model selection is a well-studied topic in parametric and nonparametric regression or density estimation, selection of possibly high-dimensional nuisance parameters in semiparametric problems is far less developed. In this paper, we propose a selective machine learning framework for making inferences about a finite-dimensional functional defined on a semiparametric model, when the latter admits a doubly robust estimating function and several candidate machine learning algorithms are available for estimating the nuisance parameters. We introduce a new selection criterion aimed at bias reduction in estimating the functional of interest based on a novel definition of pseudo risk inspired by the double robustness property. Intuitively, the proposed criterion selects a pair of learners with the smallest pseudo risk, so that the estimated functional is least sensitive to perturbations of a nuisance parameter. We establish an oracle property for a multi-fold cross-validation version of the new selection criterion that states that our empirical criterion performs nearly as well as an oracle with a priori knowledge of the pseudo risk for each pair of candidate learners. Finally, we apply the approach to model selection of a semiparametric estimator of average treatment effect given an ensemble of candidate machine learners to account for confounding in an observational study that we illustrate in simulations and a data application.

Variable selection in high dimensions for discrete-outcome individualized treatment rules: Reducing severity of depression symptoms.

Despite growing interest in estimating individualized treatment rules, little attention has been given the binary outcome setting. Estimation is challenging with nonlinear link functions, especially when variable selection is needed. We use a new computational approach to solve a recently proposed doubly robust regularized estimating equation to accomplish this difficult task in a case study of depression treatment. We demonstrate an application of this new approach in combination with a weighted and penalized estimating equation to this challenging binary outcome setting. We demonstrate the double robustness of the method and its effectiveness for variable selection. The work is motivated by and applied to an analysis of treatment for unipolar depression using a population of patients treated at Kaiser Permanente Washington.

An inverse probability weighted regression method that accounts for right-censoring for causal inference with multiple treatments and a binary outcome.

Comparative effectiveness research often involves evaluating the differences in the risks of an event of interest between two or more treatments using observational data. Often, the post-treatment outcome of interest is whether the event happens within a pre-specified time window, which leads to a binary outcome. One source of bias for estimating the causal treatment effect is the presence of confounders, which are usually controlled using propensity score-based methods. An additional source of bias is right-censoring, which occurs when the information on the outcome of interest is not completely available due to dropout, study termination, or treatment switch before the event of interest. We propose an inverse probability weighted regression-based estimator that can simultaneously handle both confounding and right-censoring, calling the method CIPWR, with the letter C highlighting the censoring component. CIPWR estimates the average treatment effects by averaging the predicted outcomes obtained from a logistic regression model that is fitted using a weighted score function. The CIPWR estimator has a double robustness property such that estimation consistency can be achieved when either the model for the outcome or the models for both treatment and censoring are correctly specified. We establish the asymptotic properties of the CIPWR estimator for conducting inference, and compare its finite sample performance with that of several alternatives through simulation studies. The methods under comparison are applied to a cohort of prostate cancer patients from an insurance claims database for comparing the adverse effects of four candidate drugs for advanced stage prostate cancer.

A Within-Group Approach to Ensemble Machine Learning Methods for Causal Inference in Multilevel Studies

Machine learning (ML) methods for causal inference have gained popularity due to their flexibility to predict the outcome model and the propensity score. In this article, we provide a within-group approach for ML-based causal inference methods in order to robustly estimate average treatment effects in multilevel studies when there is cluster-level unmeasured confounding. We focus on one particular ML-based causal inference method based on the targeted maximum likelihood estimation (TMLE) with an ensemble learner called SuperLearner. Through our simulation studies, we observe that training TMLE within groups of similar clusters helps remove bias from cluster-level unmeasured confounders. Also, using within-group propensity scores estimated from fixed effects logistic regression increases the robustness of the proposed within-group TMLE method. Even if the propensity scores are partially misspecified, the within-group TMLE still produces robust ATE estimates due to double robustness with flexible modeling, unlike parametric-based inverse propensity weighting methods. We demonstrate our proposed methods and conduct sensitivity analyses against the number of groups and individual-level unmeasured confounding to evaluate the effect of taking an eighth-grade algebra course on math achievement in the Early Childhood Longitudinal Study.

Doubly Robust Capture-Recapture Methods for Estimating Population Size

Estimation of population size using incomplete lists has a long history across many biological and social sciences. For example, human rights groups often construct partial lists of victims of armed conflicts, to estimate the total number of victims. Earlier statistical methods for this setup often use parametric assumptions, or rely on suboptimal plug-in-type nonparametric estimators; but both approaches can lead to substantial bias, the former via model misspecification and the latter via smoothing. Under an identifying assumption that two lists are conditionally independent given measured covariates, we make several contributions. First, we derive the nonparametric efficiency bound for estimating the capture probability, which indicates the best possible performance of any estimator, and sheds light on the statistical limits of capture-recapture methods. Then we present a new estimator, that has a double robustness property new to capture-recapture, and is near-optimal in a nonasymptotic sense, under relatively mild nonparametric conditions. Next, we give a confidence interval construction method for total population size from generic capture probability estimators, and prove nonasymptotic near-validity. Finally, we apply them to estimate the number of killings and disappearances in Peru during its internal armed conflict between 1980 and 2000. Supplementary materials for this article are available online.

Non-Asymptotic Bounds of AIPW Estimators for Means with Missingness at Random

The augmented inverse probability weighting is well known for its double robustness in missing data and causal inference. If either the propensity score model or the outcome regression model is correctly specified, the estimator is guaranteed to be consistent. Another important property of the augmented inverse probability weighting is that it can achieve first-order equivalence to the oracle estimator in which all nuisance parameters are known, even if the fitted models do not converge at the parametric root-n rate. We explore the non-asymptotic properties of the augmented inverse probability weighting estimator to infer the population mean with missingness at random. We also consider inferences of the mean outcomes on the observed group and on the unobserved group.