The Annals of Applied Statistics

Estimating population average causal effects in the presence of non-overlap: The effect of natural gas compressor station exposure on cancer mortality

Rachel C. Nethery, Fabrizia Mealli, and Francesca Dominici

Full-text: Access denied (no subscription detected)

We're sorry, but we are unable to provide you with the full text of this article because we are not able to identify you as a subscriber. If you have a personal subscription to this journal, then please login. If you are already logged in, then you may need to update your profile to register your subscription. Read more about accessing full-text


Most causal inference studies rely on the assumption of overlap to estimate population or sample average causal effects. When data suffer from non-overlap, estimation of these estimands requires reliance on model specifications due to poor data support. All existing methods to address non-overlap, such as trimming or down-weighting data in regions of poor data support, change the estimand so that inference cannot be made on the sample or the underlying population. In environmental health research settings where study results are often intended to influence policy, population-level inference may be critical and changes in the estimand can diminish the impact of the study results, because estimates may not be representative of effects in the population of interest to policymakers. Researchers may be willing to make additional, minimal modeling assumptions in order to preserve the ability to estimate population average causal effects. We seek to make two contributions on this topic. First, we propose a flexible, data-driven definition of propensity score overlap and non-overlap regions. Second, we develop a novel Bayesian framework to estimate population average causal effects with minor model dependence and appropriately large uncertainties in the presence of non-overlap and causal effect heterogeneity. In this approach the tasks of estimating causal effects in the overlap and non-overlap regions are delegated to two distinct models suited to the degree of data support in each region. Tree ensembles are used to nonparametrically estimate individual causal effects in the overlap region, where the data can speak for themselves. In the non-overlap region where insufficient data support means reliance on model specification is necessary, individual causal effects are estimated by extrapolating trends from the overlap region via a spline model. The promising performance of our method is demonstrated in simulations. Finally, we utilize our method to perform a novel investigation of the causal effect of natural gas compressor station exposure on cancer outcomes. Code and data to implement the method and reproduce all simulations and analyses is available on Github (

Article information

Ann. Appl. Stat., Volume 13, Number 2 (2019), 1242-1267.

Received: May 2018
Revised: November 2018
First available in Project Euclid: 17 June 2019

Permanent link to this document

Digital Object Identifier

Mathematical Reviews number (MathSciNet)

Zentralblatt MATH identifier

Overlap propensity score Bayesian additive regression trees splines natural gas cancer mortality


Nethery, Rachel C.; Mealli, Fabrizia; Dominici, Francesca. Estimating population average causal effects in the presence of non-overlap: The effect of natural gas compressor station exposure on cancer mortality. Ann. Appl. Stat. 13 (2019), no. 2, 1242--1267. doi:10.1214/18-AOAS1231.

Export citation


  • Austin, P. C. (2011). An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivar. Behav. Res. 46 399–424.
  • Bonato, V., Baladandayuthapani, V., Broom, B. M., Sulman, E. P., Aldape, K. D. and Do, K.-A. (2011). Bayesian ensemble methods for survival prediction in gene expression data. Bioinformatics 27 359–367.
  • Chipman, H. A., George, E. I. and McCulloch, R. E. (2010). BART: Bayesian additive regression trees. Ann. Appl. Stat. 4 266–298.
  • Cole, S. R. and Hernán, M. A. (2008). Constructing inverse probability weights for marginal structural models. Am. J. Epidemiol. 168 656–664.
  • Cristianini, N. and Shawe-Taylor, J. (2000). An Introduction to Support Vector Machines and Other Kernel-Based Learning Methods. Cambridge Univ. Press, Cambridge.
  • Crump, R. K., Hotz, V. J., Imbens, G. W. and Mitnik, O. A. (2009). Dealing with limited overlap in estimation of average treatment effects. Biometrika 96 187–199.
  • D’Amour, A., Deng, P., Feller, A., Lei, L. and Sekhon, J. (2017). Overlap in observational studies with high-dimensional covariates. Available at arXiv:1711.02582.
  • Di, Q., Dai, L., Wang, Y., Zanobetti, A., Choirat, C., Schwartz, J. D. and Dominici, F. (2017a). Association of short-term exposure to air pollution with mortality in older adults. JAMA 318 2446–2456.
  • Di, Q., Wang, Y., Zanobetti, A., Wang, Y., Koutrakis, P., Choirat, C., Dominici, F. and Schwartz, J. D. (2017b). Air pollution and mortality in the medicare population. N. Engl. J. Med. 376 2513–2522.
  • Finkel, M. L. (2016). Shale gas development and cancer incidence in southwest Pennsylvania. Public Health 141 198–206.
  • Golding, B. and Watson, W. (1999). Possible mechanisms of carcinogenesis after exposure to benzene. IARC Sci. Publ. 150 75–88.
  • Gutman, R. and Rubin, D. B. (2015). Estimation of causal effects of binary treatments in unconfounded studies. Stat. Med. 34 3381–3398.
  • Hahn, P. R., Murray, J. and Carvalho, C. M. (2018). Bayesian regression tree models for causal inference: Regularization, confounding, and heterogeneous effects. Preprint. Available at arXiv:1706.09523v2.
  • Hill, J. L. (2011). Bayesian nonparametric modeling for causal inference. J. Comput. Graph. Statist. 20 217–240.
  • Hill, J. and Su, Y.-S. (2013). Assessing lack of common support in causal inference using Bayesian nonparametrics: Implications for evaluating the effect of breastfeeding on children’s cognitive outcomes. Ann. Appl. Stat. 7 1386–1420.
  • Ho, D. E., Imai, K., King, G. and Stuart, E. A. (2007). Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference. Polit. Anal. 15 199–236.
  • Kassotis, C. D., Tillitt, D. E., Lin, C.-H., McElroy, J. A. and Nagel, S. C. (2016). Endocrine-disrupting chemicals and oil and natural gas operations: Potential environmental contamination and recommendations to assess complex environmental mixtures. Environ. Health Perspect. 124 256.
  • Kindo, B. P., Wang, H. and Peña, E. A. (2016). Multinomial probit Bayesian additive regression trees. Stat 5 119–131.
  • King, G. and Zeng, L. (2005). The dangers of extreme counterfactuals. Polit. Anal. 14 131–159.
  • Kloczko, N. (2015). Summary on compressor stations and health impacts.
  • Li, F., Morgan, K. L. and Zaslavsky, A. M. (2018). Balancing covariates via propensity score weighting. J. Amer. Statist. Assoc. 113 390–400.
  • Li, F. and Thomas, L. E. (2018). Addressing extreme propensity scores via the overlap weights. Am. J. Epidemiol. kwy201.
  • Linero, A. R. (2018). Bayesian regression trees for high-dimensional prediction and variable selection. J. Amer. Statist. Assoc. 113 626–636.
  • Liu, Y., Traskin, M., Lorch, S. A., George, E. I. and Small, D. (2015). Ensemble of trees approaches to risk adjustment for evaluating a hospital’s performance. Health Care Manage. Sci. 18 58–66.
  • Maltoni, C., Ciliberti, A., Cotti, G., Conti, B. and Belpoggi, F. (1989). Benzene, an experimental multipotential carcinogen: Results of the long-term bioassays performed at the Bologna Institute of Oncology. Environ. Health Perspect. 82 109.
  • McKenzie, L. M., Allshouse, W. B., Byers, T. E., Bedrick, E. J., Serdar, B. and Adgate, J. L. (2017). Childhood hematologic cancer and residential proximity to oil and gas development. PLoS ONE 12 e0170423.
  • Messersmith, D., Brockett, D. and Loveland, C. (2015). Understanding natural gas compressor stations. Penn State Extension.
  • Mokdad, A. H., Dwyer-Lindgren, L., Fitzmaurice, C., Stubbs, R. W., Bertozzi-Villa, A., Morozoff, C., Charara, R., Allen, C., Naghavi, M. et al. (2017). Trends and patterns of disparities in cancer mortality among US counties, 1980–2014. JAMA 317 388–406.
  • Nethery, R. C, Mealli, F. and Dominici, F. (2019). Supplement to “Estimating population average causal effects in the presence of non-overlap: The effect of natural gas compressor station exposure on cancer mortality.” DOI:10.1214/18-AOAS1231SUPP.
  • Oak Ridge National Laboratory (2017). Natural gas compressor stations. Accessed: 2018-03-29.
  • Pellegriti, G., Frasca, F., Regalbuto, C., Squatrito, S. and Vigneri, R. (2013). Worldwide increasing incidence of thyroid cancer: Update on epidemiology and risk factors. J. Cancer Epidemiol. 2013.
  • Pennsylvania Department of Environmental Protection (2010). Southwestern Pennsylvania Marcellus Shale short-term ambient air sampling report. Accessed: 2018-03-29.
  • Petersen, M. L., Porter, K. E., Gruber, S., Wang, Y. and van der Laan, M. J. (2012). Diagnosing and responding to violations in the positivity assumption. Stat. Methods Med. Res. 21 31–54.
  • R Core Team (2016). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.
  • Robert Wood Johnson Foundation (2014). 2014 county health rankings and roadmaps. Prepared by Social Explorer. Accessed: 2018-03-27.
  • Rosenbaum, P. R. and Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika 70 41–55.
  • Rubin, D. B. (1974). Estimating causal effects of treatments in randomized and nonrandomized studies. J. Educ. Psychol. 66 688–701.
  • Rubin, D. B. (1980). Randomization analysis of experimental data: The Fisher randomization test comment. J. Amer. Statist. Assoc. 75 591–593.
  • Schmidhuber, J. (2015). Deep learning in neural networks: An overview. Neural Netw. 61 85–117.
  • Southwest Pennsylvania Environmental Health Project (2015). Summary on compressor stations and health impacts. Accessed 2018-03-27.
  • Sparapani, R. A., Logan, B. R., McCulloch, R. E. and Laud, P. W. (2016). Nonparametric survival analysis using Bayesian Additive Regression Trees (BART). Stat. Med. 35 2741–2753.
  • US Census Bureau (2014). American community survey 2014 (5 year estimates). Prepared by Social Explorer. Accessed: 2018-03-27.
  • US EPA (2018). Endocrine disruptor screening program (EDSP) estrogen receptor bioactivity. Accessed: 2018-03-27.
  • Wang, C., Parmigiani, G. and Dominici, F. (2012). Bayesian effect estimation accounting for adjustment uncertainty. Biometrics 68 661–686.
  • Wang, C., Dominici, F., Parmigiani, G. and Zigler, C. M. (2015). Accounting for uncertainty in confounder and effect modifier selection when estimating average causal effects in generalized linear models. Biometrics 71 654–665.
  • Westreich, D. and Cole, S. R. (2010). Invited commentary: Positivity in practice. Am. J. Epidemiol. 171 674–677.
  • Westreich, D., Lessler, J. and Funk, M. J. (2010). Propensity score estimation: Neural networks, support vector machines, decision trees (CART), and meta-classifiers as alternatives to logistic regression. J. Clin. Epidemiol. 63 826–833.
  • Wolf Eagle Environmental (2009). Town of Dish, Texas ambient air monitoring analysis final report. Accessed: 2018-05-08.
  • World Trade Center Health Program (2015). Minimum latency & types or categories of cancer. Accessed: 2018-03-29.
  • Yang, S. and Ding, P. (2018). Asymptotic inference of causal effects with observational studies trimmed by the estimated propensity scores. Biometrika 105 487–493.

Supplemental materials

  • Sampling details, additional simulations, and supplementary tables and figures. Section 1 of the Supplementary Materials contains a step-by-step description of the BART${}+{}$SPL MCMC sampling scheme. Section 2 describes the data and results from simulations to test the performance of BART${}+{}$SPL for binary outcomes. Section 3 provides the supplementary tables and figures referenced in the text.