skip to main content

Propensity score methods account for selection bias in observational studies. However, the consistency of the propensity score estimators strongly depends on a correct specification of the propensity score model. Logistic regression and, with increasing popularity, machine learning tools are used to estimate propensity scores. We introduce a stacked generalization ensemble learning approach to improve propensity score estimation by fitting a meta learner on the predictions of a suitable set of diverse base learners. We perform a comprehensive Monte Carlo simulation study, implementing a broad range of scenarios that mimic characteristics of typical data sets in educational studies. The population average treatment effect is estimated using the propensity score in Inverse Probability of Treatment Weighting. Our proposed stacked ensembles, especially using gradient boosting machines as a meta learner trained on a set of 12 base learner predictions, led to superior reduction of bias compared to the current state-of-the-art in propensity score estimation. Further, our simulations imply that commonly used balance measures (averaged standardized absolute mean differences) might be misleading as propensity score model selection criteria. We apply our proposed model - which we call GBM-Stack - to assess the population average treatment effect of a Supplemental Instruction (SI) program in an introductory psychology (PSY 101) course at San Diego State University. Our analysis provides evidence that moving the whole population to SI attendance would on average lead to 1.69 times higher odds to pass the PSY 101 class compared to not offering SI, with a 95% bootstrap confidence interval of (1.31, 2.20).  more » « less
Award ID(s):
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
Journal of educational data mining
Page Range / eLocation ID:
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. 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.

    more » « less
  2. Solar flare prediction is a central problem in space weather forecasting and has captivated the attention of a wide spectrum of researchers due to recent advances in both remote sensing as well as machine learning and deep learning approaches. The experimental findings based on both machine and deep learning models reveal significant performance improvements for task specific datasets. Along with building models, the practice of deploying such models to production environments under operational settings is a more complex and often time-consuming process which is often not addressed directly in research settings. We present a set of new heuristic approaches to train and deploy an operational solar flare prediction system for ≥M1.0-class flares with two prediction modes: full-disk and active region-based. In full-disk mode, predictions are performed on full-disk line-of-sight magnetograms using deep learning models whereas in active region-based models, predictions are issued for each active region individually using multivariate time series data instances. The outputs from individual active region forecasts and full-disk predictors are combined to a final full-disk prediction result with a meta-model. We utilized an equal weighted average ensemble of two base learners’ flare probabilities as our baseline meta learner and improved the capabilities of our two base learners by training a logistic regression model. The major findings of this study are: 1) We successfully coupled two heterogeneous flare prediction models trained with different datasets and model architecture to predict a full-disk flare probability for next 24 h, 2) Our proposed ensembling model, i.e., logistic regression, improves on the predictive performance of two base learners and the baseline meta learner measured in terms of two widely used metrics True Skill Statistic (TSS) and Heidke Skill Score (HSS), and 3) Our result analysis suggests that the logistic regression-based ensemble (Meta-FP) improves on the full-disk model (base learner) by ∼9% in terms TSS and ∼10% in terms of HSS. Similarly, it improves on the AR-based model (base learner) by ∼17% and ∼20% in terms of TSS and HSS respectively. Finally, when compared to the baseline meta model, it improves on TSS by ∼10% and HSS by ∼15%. 
    more » « less
  3. For large observational studies lacking a control group (unlike randomized controlled trials, RCT), propensity scores (PS) are often the method of choice to account for pre-treatment confounding in baseline characteristics, and thereby avoid substantial bias in treatment estimation. A vast majority of PS techniques focus on average treatment effect estimation, without any clear consensus on how to account for confounders, especially in a multiple treatment setting. Furthermore, for time-to event outcomes, the analytical framework is further complicated in presence of high censoring rates (sometimes, due to non-susceptibility of study units to a disease), imbalance between treatment groups, and clustered nature of the data (where, survival outcomes appear in groups). Motivated by a right-censored kidney transplantation dataset derived from the United Network of Organ Sharing (UNOS), we investigate and compare two recent promising PS procedures, (a) the generalized boosted model (GBM), and (b) the covariate-balancing propensity score (CBPS), in an attempt to decouple the causal effects of treatments (here, study subgroups, such as hepatitis C virus (HCV) positive/negative donors, and positive/negative recipients) on time to death of kidney recipients due to kidney failure, post transplantation. For estimation, we employ a 2-step procedure which addresses various complexities observed in the UNOS database within a unified paradigm. First, to adjust for the large number of confounders on the multiple sub-groups, we fit multinomial PS models via procedures (a) and (b). In the next stage, the estimated PS is incorporated into the likelihood of a semi-parametric cure rate Cox proportional hazard frailty model via inverse probability of treatment weighting, adjusted for multi-center clustering and excess censoring, Our data analysis reveals a more informative and superior performance of the full model in terms of treatment effect estimation, over sub-models that relaxes the various features of the event time dataset. 
    more » « less
  4. Abstract Propensity score weighting is a tool for causal inference to adjust for measured confounders in observational studies. In practice, data often present complex structures, such as clustering, which make propensity score modeling and estimation challenging. In addition, for clustered data, there may be unmeasured cluster-level covariates that are related to both the treatment assignment and outcome. When such unmeasured cluster-specific confounders exist and are omitted in the propensity score model, the subsequent propensity score adjustment may be biased. In this article, we propose a calibration technique for propensity score estimation under the latent ignorable treatment assignment mechanism, i. e., the treatment-outcome relationship is unconfounded given the observed covariates and the latent cluster-specific confounders. We impose novel balance constraints which imply exact balance of the observed confounders and the unobserved cluster-level confounders between the treatment groups. We show that the proposed calibrated propensity score weighting estimator is doubly robust in that it is consistent for the average treatment effect if either the propensity score model is correctly specified or the outcome follows a linear mixed effects model. Moreover, the proposed weighting method can be combined with sampling weights for an integrated solution to handle confounding and sampling designs for causal inference with clustered survey data. In simulation studies, we show that the proposed estimator is superior to other competitors. We estimate the effect of School Body Mass Index Screening on prevalence of overweight and obesity for elementary schools in Pennsylvania. 
    more » « less
  5. Abstract Standard estimators of the global average treatment effect can be biased in the presence of interference. This paper proposes regression adjustment estimators for removing bias due to interference in Bernoulli randomized experiments. We use a fitted model to predict the counterfactual outcomes of global control and global treatment. Our work differs from standard regression adjustments in that the adjustment variables are constructed from functions of the treatment assignment vector, and that we allow the researcher to use a collection of any functions correlated with the response, turning the problem of detecting interference into a feature engineering problem. We characterize the distribution of the proposed estimator in a linear model setting and connect the results to the standard theory of regression adjustments under SUTVA. We then propose an estimator that allows for flexible machine learning estimators to be used for fitting a nonlinear interference functional form. We propose conducting statistical inference via bootstrap and resampling methods, which allow us to sidestep the complicated dependences implied by interference and instead rely on empirical covariance structures. Such variance estimation relies on an exogeneity assumption akin to the standard unconfoundedness assumption invoked in observational studies. In simulation experiments, our methods are better at debiasing estimates than existing inverse propensity weighted estimators based on neighborhood exposure modeling. We use our method to reanalyze an experiment concerning weather insurance adoption conducted on a collection of villages in rural China. 
    more » « less