Stein-like estimators for causal mediation analysis in randomized trials

Causal mediation analysis aims to estimate natural direct and natural indirect effects under clearly specified assumptions. Traditional mediation analysis based on Ordinary Least Squares assumes an absence of unmeasured causes to the putative mediator and outcome. When these assumptions cannot be justified, instrumental variable estimators can be used in order to produce an asymptotically unbiased estimator of the mediator-outcome link, commonly referred to as a Two-Stage Least Squares estimator. Such bias removal, however, comes at the cost of variance inflation. A Semi-Parametric Stein-Like estimator has been proposed in the literature that strikes a natural trade-off between the unbiasedness of the Two-Stage Least Squares procedure and the relatively small variance of the Ordinary Least Squares estimator. The Semi-Parametric Stein-Like estimator has the advantage of allowing for a direct estimation of its shrinkage parameter. In this paper, we demonstrate how this Stein-like estimator can be implemented in the context of the estimation of natural direct and natural indirect effects of treatments in randomized controlled trials. The performance of the competing methods is studied in a simulation study, in which both the strength of hidden confounding and the strength of the instruments are independently varied. These considerations are motivated by a trial in mental health, evaluating the impact of a primary care-based intervention to reduce depression in the elderly.

[1]  Cedric E. Ginestet,et al.  Dose–response modeling in mental health using stein‐like estimators with instrumental variables , 2017, Statistics in medicine.

[2]  G. Judge,et al.  A Minimum Mean Squared Error Semiparametric Combining Estimator , 2012 .

[3]  A Risk Superior Semiparametric Estimator for Overidentified Linear Models , 2012 .

[4]  Marshall M Joffe,et al.  A review of causal estimation of effects in mediation analyses , 2012, Statistical methods in medical research.

[5]  G. Judge,et al.  An Information Theoretic Approach to Econometrics , 2011 .

[6]  Dylan S. Small.  Mediation Analysis Without Sequential Ignorability: Using Baseline Covariates Interacted with Random Assignment as Instrumental Variables , 2011, 1109.1070.

[7]  Frank Windmeijer,et al.  Identification of causal effects on binary outcomes using structural mean models , 2010, Biostatistics.

[8]  Richard Emsley,et al.  Ian R White interventions Mediation and moderation of treatment effects in randomised controlled trials of complex , 2010 .

[9]  L. Keele,et al.  Identification, Inference and Sensitivity Analysis for Causal Mediation Effects , 2010, 1011.1079.

[10]  Frank Windmeijer,et al.  Instrumental Variable Estimators for Binary Outcomes , 2009 .

[11]  T. VanderWeele.  Simple relations between principal stratification and direct and indirect effects , 2008 .

[12]  Jeffrey M Albert,et al.  Mediation analysis via potential outcomes models , 2008, Statistics in medicine.

[13]  Mark Cary,et al.  Causal mediation analyses for randomized trials , 2008, Health Services and Outcomes Research Methodology.

[14]  D. Mackinnon.  Introduction to Statistical Mediation Analysis , 2008 .

[15]  Graham Dunn,et al.  Modelling treatment‐effect heterogeneity in randomized controlled trials of complex interventions (psychological treatments) , 2007, Statistics in medicine.

[16]  Kevin G Lynch,et al.  Causal Mediation Analyses with Rank Preserving Models , 2007, Biometrics.

[17]  G. Judge,et al.  Combining estimators to improve structural model estimation and inference under quadratic loss , 2005 .

[18]  D. Rubin.  Direct and Indirect Causal Effects via Potential Outcomes * , 2004 .

[19]  Thomas R Ten Have,et al.  Reducing suicidal ideation and depressive symptoms in depressed older primary care patients: a randomized controlled trial. , 2004, JAMA.

[20]  G. Judge,et al.  A Semiparametric Basis for Combining Estimation Problems Under Quadratic Loss , 2004 .

[21]  Jeffrey M. Wooldridge,et al.  Solutions Manual and Supplementary Materials for Econometric Analysis of Cross Section and Panel Data , 2003 .

[22]  Judea Pearl,et al.  Direct and Indirect Effects , 2001, UAI.

[23]  J. Angrist,et al.  Jackknife Instrumental Variables Estimation , 1995 .

[24]  J. Robins,et al.  Identifiability and Exchangeability for Direct and Indirect Effects , 1992, Epidemiology.

[25]  C. Stein,et al.  Estimation with Quadratic Loss , 1992 .

[26]  J. Berger.  Statistical Decision Theory and Bayesian Analysis , 1988 .

[27]  D. A. Kenny,et al.  The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. , 1986, Journal of personality and social psychology.

[28]  Arnold Zellner,et al.  Minimum Expected Loss (MELO) Estimators for Functions of Parameters and Structural Coefficients of Econometric Models , 1979 .

[29]  Arnold Zellner,et al.  Estimation of functions of population means and regression coefficients including structural coefficients : A minimum expected loss (MELO) approach , 1978 .

[30]  B. Efron,et al.  Stein's Estimation Rule and Its Competitors- An Empirical Bayes Approach , 1973 .