Causal Inference Studies Research Articles

A Bayesian Semiparametric Approach to Intermediate Variables in Causal Inference

In causal inference studies, treatment comparisons often need to be adjusted for confounded post-treatment variables. Principal stratification (PS) is a framework to deal with such variables within the potential outcome approach to causal inference. Continuous intermediate variables introduce inferential challenges to PS analysis. Existing methods either dichotomize the intermediate variable, or assume a fully parametric model for the joint distribution of the potential intermediate variables. However, the former is subject to information loss and arbitrary choice of the cutoff point and the latter is often inadequate to represent complex distributional and clustering features. We propose a Bayesian semiparametric approach that consists of a flexible parametric model for the potential outcomes and a Bayesian nonparametric model for the potential intermediate outcomes using a Dirichlet process mixture (DPM) model. The DPM approach provides flexibility in modeling the possibly complex joint distribution of the potential intermediate outcomes and offers better interpretability of results through its clustering feature. Gibbs sampling based posterior inference is developed. We illustrate the method by two applications: one concerning partial compliance in a randomized clinical trial, and one concerning the causal mechanism between physical activity, body mass index, and cardiovascular disease in the observational Swedish National March Cohort study.

The association between total urinary arsenic concentration and renal dysfunction in a community-based population from central Taiwan

Arsenic (As) is an important environmental toxicant that can cause cancer and cardiovascular disease, but the relationship between As exposure and renal dysfunction is not clear. The aim of this study is to examine the association between As exposure and renal dysfunction in a community-based population in central Taiwan. One thousand and forty-three subjects were recruited between 2002 and 2005. The risk for type 2 diabetes was increased by 2-fold (p<0.05) in subjects with total urinary As (U-As)>75μgg−1 creatinine as compared with subjects whose U-As was ⩽35μgg−1 creatinine after the adjustment for potential confounders. The adjusted odds ratio for an abnormal β2 microglobulin (B2MG>0.154mgL−1) was significantly higher in subjects with U-As>35μgg−1 creatinine as compared with the reference group adjusted for age, sex, living area, cigarette smoking, diabetes, and hypertension. The risk for abnormal B2MG and estimated glomerular filtration rate (eGFR<90mL min−1(1.73m2)−1) was both increased around 2-fold (p<0.05) in subjects with U-As>75μgg−1 creatinine as compared with those with U-As⩽35μgg−1 creatinine adjusted for all the risk factors plus lead (Pb), cadmium and nickel. The prevalence of abnormal B2MG was 4.82 times higher in subjects with both over the median levels of U-As (85.1μgL−1) and urinary Pb (18.9μgL−1) as compared to both lower than the median (p<0.001). These results indicate that U-As might relate to renal dysfunction even other important risk factors were taken into account. Follow-up studies for causal inference are warranted.

Causal Inference in Multisensory Perception

Perceptual events derive their significance to an animal from their meaning about the world, that is from the information they carry about their causes. The brain should thus be able to efficiently infer the causes underlying our sensory events. Here we use multisensory cue combination to study causal inference in perception. We formulate an ideal-observer model that infers whether two sensory cues originate from the same location and that also estimates their location(s). This model accurately predicts the nonlinear integration of cues by human subjects in two auditory-visual localization tasks. The results show that indeed humans can efficiently infer the causal structure as well as the location of causes. By combining insights from the study of causal inference with the ideal-observer approach to sensory cue combination, we show that the capacity to infer causal structure is not limited to conscious, high-level cognition; it is also performed continually and effortlessly in perception.

Causation in risk assessment and management: models, inference, biases, and a microbial risk–benefit case study

Causal inference of exposure–response relations from data is a challenging aspect of risk assessment with important implications for public and private risk management. Such inference, which is fundamentally empirical and based on exposure (or dose)–response models, seldom arises from a single set of data; rather, it requires integrating heterogeneous information from diverse sources and disciplines including epidemiology, toxicology, and cell and molecular biology. The causal aspects we discuss focus on these three aspects: • Drawing sound inferences about causal relations from one or more observational studies; • Addressing and resolving biases that can affect a single multivariate empirical exposure–response study; and • Applying the results from these considerations to the microbiological risk management of human health risks and benefits of a ban on antibiotic use in animals, in the context of banning enrofloxacin or macrolides, antibiotics used against bacterial illnesses in poultry, and the effects of such bans on changing the risk of human food-borne campylobacteriosis infections. The purposes of this paper are to describe novel causal methods for assessing empirical causation and inference; exemplify how to deal with biases that routinely arise in multivariate exposure– or dose–response modeling; and provide a simplified discussion of a case study of causal inference using microbial risk analysis as an example. The case study supports the conclusion that the human health benefits from a ban are unlikely to be greater than the excess human health risks that it could create, even when accounting for uncertainty. We conclude that quantitative causal analysis of risks is a preferable to qualitative assessments because it does not involve unjustified loss of information and is sound under the inferential use of risk results by management.

WHICH COMES FIRST, CAUSE OR EFFECT?

ABSTRACTThis is a short introduction to Rubin's model for the study of causal inference in experiments and observational studies. More details, proofs of mathematical assertions, and expanded discussions can be found in the references.