Tools For Causal Inference Research Articles

An Information Flow-Based Sea Surface Height Reconstruction Through Machine Learning

The advent of satellite altimetry datasets of sea surface height (SSH) is a major advance in oceanography and other Earth system sciences. However, while the along-track data coverage is dense, the relatively poor resolution between tracks poses a challenge to the reconstruction of those processes such as mesoscale and submesoscale eddies. This study proposes a machine learning algorithm based on a causal inference tool, i.e., the Liang–Kleeman information flow (L-K IF) analysis, to address the challenge. For a region in the South China Sea where eddies frequently appear but unobserved, it is shown that the algorithm can reconstruct the desired mesoscale eddies in a remarkably successful way in geometry, orientation, strength, etc., while with the objective analysis interpolation or the traditional neural network technique, the results are not satisfactory. This study provides prospects for developing the next generation of SSH products with the available altimetry data.

Beyond the Mean: A Flexible Framework for Studying Causal Effects Using Linear Models

Graph-based causal models are a flexible tool for causal inference from observational data. In this paper, we develop a comprehensive framework to define, identify, and estimate a broad class of causal quantities in linearly parametrized graph-based models. The proposed method extends the literature, which mainly focuses on causal effects on the mean level and the variance of an outcome variable. For example, we show how to compute the probability that an outcome variable realizes within a target range of values given an intervention, a causal quantity we refer to as the probability of treatment success. We link graph-based causal quantities defined via the do-operator to parameters of the model implied distribution of the observed variables using so-called causal effect functions. Based on these causal effect functions, we propose estimators for causal quantities and show that these estimators are consistent and converge at a rate of N^{-1/2} under standard assumptions. Thus, causal quantities can be estimated based on sample sizes that are typically available in the social and behavioral sciences. In case of maximum likelihood estimation, the estimators are asymptotically efficient. We illustrate the proposed method with an example based on empirical data, placing special emphasis on the difference between the interventional and conditional distribution.

Using Propensity Scores for Causal Inference: Pitfalls and Tips.

Methods based on propensity score (PS) have become increasingly popular as a tool for causal inference. A better understanding of the relative advantages and disadvantages of the alternative analytic approaches can contribute to the optimal choice and use of a specific PS method over other methods. In this article, we provide an accessible overview of causal inference from observational data and two major PS-based methods (matching and inverse probability weighting), focusing on the underlying assumptions and decision-making processes. We then discuss common pitfalls and tips for applying the PS methods to empirical research and compare the conventional multivariable outcome regression and the two alternative PS-based methods (ie, matching and inverse probability weighting) and discuss their similarities and differences. Although we note subtle differences in causal identification assumptions, we highlight that the methods are distinct primarily in terms of the statistical modeling assumptions involved and the target population for which exposure effects are being estimated.

Connecting Instrumental Variable methods for causal inference to the Estimand Framework.

Causal inference methods are gaining increasing prominence in pharmaceutical drug development in light of the recently published addendum on estimands and sensitivity analysis in clinical trials to the E9 guideline of the International Council for Harmonisation. The E9 addendum emphasises the need to account for post-randomization or 'intercurrent' events that can potentially influence the interpretation of a treatment effect estimate at a trial's conclusion. Instrumental Variables (IV) methods have been used extensively in economics, epidemiology, and academic clinical studies for 'causal inference,' but less so in the pharmaceutical industry setting until now. In this tutorial article we review the basic tools for causal inference, including graphical diagrams and potential outcomes, as well as several conceptual frameworks that an IV analysis can sit within. We discuss in detail how to map these approaches to the Treatment Policy, Principal Stratum and Hypothetical 'estimand strategies' introduced in the E9 addendum, and provide details of their implementation using standard regression models. Specific attention is given to discussing the assumptions each estimation strategy relies on in order to be consistent, the extent to which they can be empirically tested and sensitivity analyses in which specific assumptions can be relaxed. We finish by applying the methods described to simulated data closely matching two recent pharmaceutical trials to further motivate and clarify the ideas.

Causal inference in suicide research: When you should (and should not!) control for extraneous variables.

Although causal inference is often straightforward in experimental contexts, few research questions in suicide are amenable to experimental manipulation and randomized control. Instead, suicide prevention specialists must rely on observational data and statistical control of confounding variables to make effective causal inferences. We provide a brief summary of recent covariate practice and a tutorial on casual inference tools for covariate selection in suicide research. We provide an introduction to modern causal inference tools, suggestions for statistical control selection, and demonstrations using simulated data. Statistical controls are often mistakenly selected due to their significant correlation with other study variables, their consistency with previous research, or no explicit reason at all. We clarify what it means to control for a variable and when controlling for the wrong covariates systematically distorts results. We describe directed acyclic graphs (DAGs) and tools for identifying the right choice of covariates. Finally, we provide four best practices for integrating causal inference tools in future studies. The use of causal model tools, such as DAGs, allows researchers to carefully and thoughtfully select statistical controls and avoid presenting distorted findings; however, limitations of this approach are discussed.

Assessment of Attitude, Subjective Norm and Perceived Behavioral Control on Physical Activity of Alcoholics Using Structural Equation Modeling

Background: Structural equation modeling (SEM) is a series of statistical methods that allow complex relationships between one or more independent variables and one or more dependent variables. Objective: The study aimed at determining the practice of physical activity using structural equation modeling as statistical approach. Methods: This cross-sectional study was conducted from August to November 2018 amongst recuperating alcoholics receiving rehabilitation in Asumbi treatment center of Homabay County, Kenya. Structural equation modeling determined the evidence of physical activity amongst recuperating alcoholics. Results: Structural model parameter estimation indicated that attitude was a better predictor (β=0.62, p<0.001, n=207), followed by subjective norm (β=0.60, p<0.001, n=207) then perceived behavioral control (β=0.55, p<0.001, n=207) was indirectly and directly predicted. Conclusion: Structural Equation Modeling is a powerful tool for causal inference among the observed and latent variables.

Computational Tools for Causal Inference in Genetics.

The advent of large-scale, phenotypically rich, and readily accessible data provides an unprecedented opportunity for epidemiologists, statistical geneticists, bioinformaticians, and also behavioral and social scientists to investigate the causes and consequences of disease. Computational tools and resources are an integral component of such endeavors, which will become increasingly important as these data continue to grow exponentially. In this review, we have provided an overview of computational software and databases that have been developed to assist with analyses in causal inference. This includes online tools that can be used to help generate hypotheses, publicly accessible resources that store summary-level information for millions of genetic markers, and computational approaches that can be used to leverage this wealth of data to study causal relationships.

Mendelian randomization as a tool for causal inference in human nutrition and metabolism.

The current review describes the fundamentals of the Mendelian randomization framework and its current application for causal inference in human nutrition and metabolism. In the Mendelian randomization framework, genetic variants that are strongly associated with the potential risk factor are used as instrumental variables to determine whether the risk factor is a cause of the disease. Mendelian randomization studies are less susceptible to confounding and reverse causality compared with traditional observational studies. The Mendelian randomization study design has been increasingly used in recent years to appraise the causal associations of various nutritional factors, such as milk and alcohol intake, circulating levels of micronutrients and metabolites, and obesity with risk of different health outcomes. Mendelian randomization studies have confirmed some but challenged other nutrition-disease associations recognized by traditional observational studies. Yet, the causal role of many nutritional factors and intermediate metabolic changes for health and disease remains unresolved. Mendelian randomization can be used as a tool to improve causal inference in observational studies assessing the role of nutritional factors and metabolites in health and disease. There is a need for more large-scale genome-wide association studies to identify more genetic variants for nutritional factors that can be utilized for Mendelian randomization analyses.

Do-search: A Tool for Causal Inference and Study Design with Multiple Data Sources.

Epidemiologic evidence is based on multiple data sources including clinical trials, cohort studies, surveys, registries, and expert opinions. Merging information from different sources opens up new possibilities for the estimation of causal effects. We show how causal effects can be identified and estimated by combining experiments and observations in real and realistic scenarios. As a new tool, we present do-search, a recently developed algorithmic approach that can determine the identifiability of a causal effect. The approach is based on do-calculus, and it can utilize data with nontrivial missing data and selection bias mechanisms. When the effect is identifiable, do-search outputs an identifying formula on which numerical estimation can be based. When the effect is not identifiable, we can use do-search to recognize additional data sources and assumptions that would make the effect identifiable. Throughout the article, we consider the effect of salt-adding behavior on blood pressure mediated by the salt intake as an example. The identifiability of this effect is resolved in various scenarios with different assumptions on confounding. There are scenarios where the causal effect is identifiable from a chain of experiments but not from survey data, as well as scenarios where the opposite is true. As an illustration, we use survey data from the National Health and Nutrition Examination Survey 2013-2016 and the results from a meta-analysis of randomized controlled trials and estimate the reduction in average systolic blood pressure under an intervention where the use of table salt is discontinued.

Phosphoproteomic quantitation and causal analysis reveal pathways in GPVI/ITAM-mediated platelet activation programs

AbstractPlatelets engage cues of pending vascular injury through coordinated adhesion, secretion, and aggregation responses. These rapid, progressive changes in platelet form and function are orchestrated downstream of specific receptors on the platelet surface and through intracellular signaling mechanisms that remain systematically undefined. This study brings together cell physiological and phosphoproteomics methods to profile signaling mechanisms downstream of the immunotyrosine activation motif (ITAM) platelet collagen receptor GPVI. Peptide tandem mass tag (TMT) labeling, sample multiplexing, synchronous precursor selection (SPS), and triple stage tandem mass spectrometry (MS3) detected >3000 significant (false discovery rate < 0.05) phosphorylation events on >1300 proteins over conditions initiating and progressing GPVI-mediated platelet activation. With literature-guided causal inference tools, >300 site-specific signaling relations were mapped from phosphoproteomics data among key and emerging GPVI effectors (ie, FcRγ, Syk, PLCγ2, PKCδ, DAPP1). Through signaling validation studies and functional screening, other less-characterized targets were also considered within the context of GPVI/ITAM pathways, including Ras/MAPK axis proteins (ie, KSR1, SOS1, STAT1, Hsp27). Highly regulated GPVI/ITAM targets out of context of curated knowledge were also illuminated, including a system of >40 Rab GTPases and associated regulatory proteins, where GPVI-mediated Rab7 S72 phosphorylation and endolysosomal maturation were blocked by TAK1 inhibition. In addition to serving as a model for generating and testing hypotheses from omics datasets, this study puts forth a means to identify hemostatic effectors, biomarkers, and therapeutic targets relevant to thrombosis, vascular inflammation, and other platelet-associated disease states.

Bayesian network analysis incorporating genetic anchors complements conventional Mendelian randomization approaches for exploratory analysis of causal relationships in complex data.

Mendelian randomization (MR) implemented through instrumental variables analysis is an increasingly popular causal inference tool used in genetic epidemiology. But it can have limitations for evaluating simultaneous causal relationships in complex data sets that include, for example, multiple genetic predictors and multiple potential risk factors associated with the same genetic variant. Here we use real and simulated data to investigate Bayesian network analysis (BN) with the incorporation of directed arcs, representing genetic anchors, as an alternative approach. A Bayesian network describes the conditional dependencies/independencies of variables using a graphical model (a directed acyclic graph) with an accompanying joint probability. In real data, we found BN could be used to infer simultaneous causal relationships that confirmed the individual causal relationships suggested by bi-directional MR, while allowing for the existence of potential horizontal pleiotropy (that would violate MR assumptions). In simulated data, BN with two directional anchors (mimicking genetic instruments) had greater power for a fixed type 1 error than bi-directional MR, while BN with a single directional anchor performed better than or as well as bi-directional MR. Both BN and MR could be adversely affected by violations of their underlying assumptions (such as genetic confounding due to unmeasured horizontal pleiotropy). BN with no directional anchor generated inference that was no better than by chance, emphasizing the importance of directional anchors in BN (as in MR). Under highly pleiotropic simulated scenarios, BN outperformed both MR (and its recent extensions) and two recently-proposed alternative approaches: a multi-SNP mediation intersection-union test (SMUT) and a latent causal variable (LCV) test. We conclude that BN incorporating genetic anchors is a useful complementary method to conventional MR for exploring causal relationships in complex data sets such as those generated from modern “omics” technologies.

The Trouble with Coarsened Exact Matching

“Balancing” methods, using matching or reweighting to improve the balance between treated and control units, have become central methodological tools for causal inference in the social sciences using cross-sectional observational data. We address here one method which has attained substantial popularity in political science, Coarsened Exact Matching (CEM) (Iacus, King, and Porro 2012). We report evidence that CEM performs substantially worse than an array of other methods and explain why it does so. We replicate five recent papers that use CEM and compare CEM-based results to those from other balancing methods. CEM drops substantially more observations, does so in non-obvious ways, can severely misidentify average treatment effects, and is much less precise than other methods. Our advice: never use CEM as the sole balancing method, and there is little to be said for using it at all. The Online Appendix for this paper is available at http://ssrn.com/abstract=3705007.

Testing Complex Social Theories With Causal Mediation Analysis and G-Computation: Toward a Better Way to Do Causal Structural Equation Modeling

Complex social scientific theories are conventionally tested using linear structural equation modeling (SEM). However, the underlying assumptions of linear SEM often prove unrealistic, making the decomposition of direct and indirect effects problematic. Recent advancements in causal mediation analysis can help to address these shortcomings, allowing for causal inference when a new set of identifying assumptions are satisfied. This article reviews how these ideas can be generalized to multiple mediators, with a focus on the posttreatment confounding and causal ordering cases. Using the potential outcome framework as a rigorous tool for causal inference, the application is the theory of procedural justice policing. Analysis of data from two randomized experiments shows that making similar parametric assumptions to SEMs and using G-computation improve the viability of effect decomposition. The article concludes with a discussion of how causal mediation analysis improves upon SEM and the potential limitation of the methods.

Tools for causal inference from cross-sectional innovation surveys with continuous or discrete variables: Theory and applications

This paper presents a new statistical toolkit by applying three techniques for data-driven causal inference from the machine learning community that are little-known among economists and innovation scholars: a conditional independence-based approach, additive noise models, and non-algorithmic inference by hand. We include three applications to CIS data to investigate public funding schemes for R&amp;D investment, information sources for innovation, and innovation expenditures and firm growth. Preliminary results provide causal interpretations of some previously-observed correlations. Our statistical 'toolkit' could be a useful complement to existing techniques.

Directed Acyclic Graphs: Alternative tool for causal inference in epidemiology and biostatistics research and teaching

The issue of causation is one of the major challenges for epidemiologists who aim to understand the association between an exposure and an outcome to explain disease patterns and potentially provide a basis for intervention. Suitably designed experimental studies can offer robust evidence of the causal relationships. The experimental studies, however, are not popular, difficult or even unethical and impossible to conduct; it would be desirable if there is a methodology for reducing bias or strengthening the causal inferences drawn from observational studies. The traditional approach of estimating causal effects in such studies is to adjust for a set of variables judged to be confounders by including them in a multiple regression. However, which variables should be adjusted for as confounders in a regression model has long been a controversial issue in epidemiology. From my observation, the adjustments using only “statistical artifacts” methods such as the p-value&lt;0.2 in univariate analysis, stepwise (forward/backward) are widely used in research and teaching in Epidemiology and Statistics but without appropriated notice on the biological or clinical relationships between exposure and outcome which may induce the bias in estimating causal effects. In this mini-review, we introduce an interesting method, namely Directed Acyclic Graphs (DAGs), which can be used to reduce the bias in estimating causal effects; it is also a good application for Epidemiology and Biostatistics teaching.

Propensity Score Weighting for Causal Inference with Clustered Data

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.

Causal Inference Using Mediation Analysis or Instrumental Variables - Full Mediation in the Absence of Conditional Independence

Both instrumental variable (IV) estimation and mediation analysis are tools for causal inference. However, IV estimation has mostly developed in economics for causal inference from observational data. In contrast, mediation analysis has mostly developed in psychology, as a tool to empirically establish the process by which an experimental manipulation brings about its effect on the dependent variable of interest. As a consequence, many researchers well versed in IV estimation are not familiar with mediation analysis, and vice versa. In this paper, we discuss the communalities and differences between IV estimation and mediation analysis. We highlight that IV estimation leverages an a priori assumption of full mediation for causal inference. In contrast, modern practice in mediation analysis focusses on testing the statistical significance of the indirect effect without too much regard for the specification of the estimated model. A drawback of this approach is that inferring mediation from the statistical significance of a (putative) indirect effect through the hypothesized mediator may be spurious altogether. We discuss specification issues and how they relate to inference about mediation, and specifically the distinction between full and partial mediation. Based on this discussion we argue in favor of further developing tests that are more diagnostic about the underlying causal structure, motivated by the implication that full mediation could be more common than currently believed.

Causal Inference Using Mediation Analysis or Instrumental Variables - Full Mediation in the Absence of Conditional Independence

Both instrumental variable (IV) estimation and mediation analysis are tools for causal inference. However, IV estimation has mostly developed in economics for causal inference from observational data. In contrast, mediation analysis has mostly developed in psychology, as a tool to empirically establish the process by which an experimental manipulation brings about its effect on the dependent variable of interest. As a consequence, many researchers well versed in IV estimation are not familiar with mediation analysis, and vice versa. In this paper, we discuss the communalities and differences between IV estimation and mediation analysis. We highlight that IV estimation leverages an a priori assumption of full mediation for causal inference. In contrast, modern practice in mediation analysis focusses on testing the statistical significance of the indirect effect without too much regard for the specification of the estimated model. A drawback of this approach is that inferring mediation from the statistical significance of a (putative) indirect effect through the hypothesized mediator may be spurious altogether. We discuss specification issues and how they relate to inference about mediation, and specifically the distinction between full and partial mediation. Based on this discussion we argue in favor of further developing tests that are more diagnostic about the underlying causal structure, motivated by the implication that full mediation could be more common than currently believed.

Graphical Models of Causation (WITHDRAWN)

The computer science and artificial intelligence literature provides powerful tools for causal inference with observational data based on graphical models of causation. A paper by Durand and Vaara ...

Standardization as a Tool for Causal Inference in Medical Research.

Traditional standardization methods have been used in medical research for a long time to standardize the effect of interest for one confounder such as age. Model-based standardization extension of these methods is used when we have more than one variable produces an effect which is the population average and has marginal causal interpretation. In this paper, we discuss the most traditional model-based standardization methods that are used to estimate the marginal causal effect of exposure. We applied these methods to data from Tehran Thyroid Study and estimated the standardized effect of exposure on outcome. Based on the simulation studies, covariate standardization is preferred except when 1) we have enough information about the mechanism of exposure or 2) the outcome is rare and exposure is frequent, so propensity score standardization is suggested.