Mediation analysis via potential outcomes models

This paper develops a causal or manipulation model framework for mediation analysis based on the concept of potential outcome. Using this framework, we provide new definitions and measures of mediation. Effects of manipulations are modeled via the linear structural model. Corresponding structural equation models (SEMs), in conjunction with two-stage least-squares estimation and the delta method, are used to perform inference. The methods are applied to data from a study of nursing interventions for postoperative pain. We address the cases of more than two treatment groups, and an interaction among mediators. For the latter, a sensitivity analysis approach to handle unidentified parameters is described. Interpretative advantages of the potential outcomes framework for mediation are emphasized.

[1]  A. Dawid.  Causal Inference without Counterfactuals , 2000 .

[2]  D. Rubin,et al.  Principal Stratification in Causal Inference , 2002, Biometrics.

[3]  Jeremy M G Taylor,et al.  A Measure of the Proportion of Treatment Effect Explained by a Surrogate Marker , 2002, Biometrics.

[4]  G Molenberghs,et al.  Evaluation of surrogate endpoints in randomized experiments with mixed discrete and continuous outcomes , 2001, Statistics in medicine.

[5]  Stephen G. Donald,et al.  Choosing the Number of Instruments , 2001 .

[6]  Rodolphe Thiébaut,et al.  Counterfactual Links to the Proportion of Treatment Effect Explained by a Surrogate Marker , 2005, Biometrics.

[7]  P Flandre,et al.  Estimating the proportion of treatment effect explained by a surrogate marker. , 1999, Statistics in Medicine.

[8]  H. Iyer,et al.  Unit–Treatment Interaction and Its Practical Consequences , 2000, Biometrics.

[9]  Niels Keiding,et al.  The Mediation Proportion: A Structural Equation Approach for Estimating the Proportion of Exposure Effect on Outcome Explained by an Intermediate Variable , 2005, Epidemiology.

[10]  E. Arjas,et al.  Causal Reasoning from Longitudinal Data * , 2004 .

[11]  J. Robins,et al.  Adjusting for differential rates of prophylaxis therapy for PCP in high- versus low-dose AZT treatment arms in an AIDS randomized trial , 1994 .

[12]  S. T. Buckland,et al.  An Introduction to the Bootstrap. , 1994 .

[13]  S L Zeger,et al.  The Evaluation of Multiple Surrogate Endpoints , 2001, Biometrics.

[14]  E. Mascha,et al.  Properties Of Bound Estimators On Treatment Effect Heterogeneity For Binary Outcomes , 2006 .

[15]  T. Lancaster,et al.  Instrumental variables and inverse probability weighting for causal inference from longitudinal observational studies , 2004, Statistical methods in medical research.

[16]  Marshall M Joffe,et al.  Weighting in instrumental variables and G‐estimation , 2003, Statistics in medicine.

[17]  Kenneth A. Bollen,et al.  Structural Equations with Latent Variables , 1989 .

[18]  E Goetghebeur,et al.  Practical properties of some structural mean analyses of the effect of compliance in randomized trials. , 1997, Controlled clinical trials.

[19]  Jeffrey M Albert,et al.  Assessing Treatment Effect Heterogeneity in Clinical Trials with Blocked Binary Outcomes , 2005, Biometrical journal. Biometrische Zeitschrift.

[20]  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.

[21]  D. A. Kenny,et al.  Process Analysis , 1981 .

[22]  B. Graubard,et al.  Statistical validation of intermediate endpoints for chronic diseases. , 1992, Statistics in medicine.

[23]  James M. Robins,et al.  Marginal Structural Models versus Structural nested Models as Tools for Causal inference , 2000 .

[24]  Bin Huang,et al.  Statistical assessment of the contribution of a mediator to an exposure outcome process. , 2001 .

[25]  D. DeMets,et al.  Surrogate End Points in Clinical Trials: Are We Being Misled? , 1996, Annals of Internal Medicine.

[26]  D. Rubin.  Statistics and Causal Inference: Comment: Which Ifs Have Causal Answers , 1986 .

[27]  P. Holland.  Statistics and Causal Inference , 1985 .

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

[29]  David A. Jaeger,et al.  Problems with Instrumental Variables Estimation when the Correlation between the Instruments and the Endogenous Explanatory Variable is Weak , 1995 .

[30]  D. Rubin.  Estimating causal effects of treatments in randomized and nonrandomized studies. , 1974 .

[31]  M Good,et al.  Sensation and Distress of Pain Scales: Reliability, Validity, and Sensitivity , 2001, Journal of Nursing Measurement.

[32]  B. Muthén.  A general structural equation model with dichotomous, ordered categorical, and continuous latent variable indicators , 1984 .

[33]  Yasuo Amemiya,et al.  A method of moments technique for fitting interaction effects in structural equation models. , 2003, The British journal of mathematical and statistical psychology.

[34]  M. Sobel.  Direct and Indirect Effects in Linear Structural Equation Models , 1987 .

[35]  J. Robins,et al.  Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. , 2000, Epidemiology.

[36]  Kajal Lahiri,et al.  Further consequences of viewing LIML as an iterated Aitken estimator , 2000 .

[37]  D. Lindley.  Seeing and Doing: the Concept of Causation , 2002 .

[38]  S. West,et al.  A comparison of methods to test mediation and other intervening variable effects. , 2002, Psychological methods.

[39]  Jeffrey M. Albert,et al.  Individual treatment effects in randomized trials with binary outcomes , 2004 .

[40]  Richard F MacLehose,et al.  The mediation proportion. , 2005, Epidemiology.

[41]  Roger L. Brown.  Assessing specific mediational effects in complex theoretical models , 1997 .