Minimax Regret Treatment Choice with Limited Validity of Experiments or with Covariates

This paper continues the investigation of minimax regret treatment choice initiated by Manski (2004). Consider a decision maker who must assign treatment to future subjects after observing outcomes experienced in a finite sample. A certain scoring rule is known to achieve minimax regret in numerous variants of this decision problem. I investigate the sensitivity of these findings to perturbations of the decision environment in realistic directions. They are: (i) The experiment may have limited validity, either because of selective noncompliance and realted problems or because the sampling universe is a potentially selective (in one of several ways) subset of the treatment population, thus samples generate potentially misleading signals even in the limit. These problems are formalized via a “bounds” approach that turns the problem into one of partial identification. (ii) Treatment outcomes may be influenced by a covariate whose effect on outcome distributions is bounded (in one of numerous probability metrics). This is interesting because introduction of a covariate with unrestricted effects leads to a pathological result. In both scenarios, small but positive perturbations leave the minimax regret decision rule unchanged (with caveats in one case). Thus, minimax regret analysis is not knife-edge dependent on ignoring certain aspects of realistic decision problems. Indeed, it recommends to entirely disregard covariates whose effect is believed to be positive but small, as well as small enough amounts of missing data. All findings are finite sample results derived by game theoretic analysis.

[1]  Karl H. Schlag,et al.  Minimax Regret in Practice-Four Examples on Treatment Choice 1 , 2006 .

[2]  Rajeev Dehejia,et al.  Program Evaluation as a Decision Problem , 1999 .

[3]  Karl H. Schlag,et al.  How to Minimize Maximum Regret in Repeated Decision-Making , 2004 .

[4]  C. Manski Statistical treatment rules for heterogeneous populations , 2003 .

[5]  C. Manski Anatomy of the Selection Problem , 1989 .

[6]  Jörg Stoye,et al.  Minimax regret treatment choice with finite samples , 2009 .

[7]  Charles F. Manski,et al.  Search Profiling with Partial Knowledge of Deterrence , 2005 .

[8]  On Some Nonparametric Selection Procedures , 1990 .

[9]  C. Manski,et al.  Competitive Lending with Partial Knowledge of Loan Repayment , 2008 .

[10]  Charles F. Manski,et al.  Minimax-regret treatment choice with missing outcome data , 2007 .

[11]  William A. Brock,et al.  Profiling Problems with Partially Identified Structure , 2006 .

[12]  C. Manski,et al.  Admissible treatment rules for a risk-averse planner with experimental data on an innovation , 2007 .

[13]  Charles F. Manski,et al.  Identification problems and decisions under ambiguity: Empirical analysis of treatment response and normative analysis of treatment choice , 2000 .

[14]  Charles F. Manski,et al.  Identification for Prediction and Decision , 2008 .

[15]  E. Vytlacil Independence, Monotonicity, and Latent Index Models: An Equivalence Result , 2002 .

[16]  Dirk Bergemann,et al.  Robust Monopoly Pricing , 2008, J. Econ. Theory.

[17]  Abraham Wald,et al.  Statistical Decision Functions , 1951 .

[18]  J. Worrall Evidence in Medicine and Evidence‐Based Medicine , 2007 .

[19]  J. Pearl,et al.  Bounds on Treatment Effects from Studies with Imperfect Compliance , 1997 .

[20]  A. Wald Statistical Decision Functions Which Minimize the Maximum Risk , 1945 .

[21]  Gary Chamberlain,et al.  BAYESIAN ASPECTS OF TREATMENT CHOICE , 2011 .

[22]  I. Glicksberg A FURTHER GENERALIZATION OF THE KAKUTANI FIXED POINT THEOREM, WITH APPLICATION TO NASH EQUILIBRIUM POINTS , 1952 .

[23]  Jörg Stoye MINIMAX REGRET TREATMENT CHOICE WITH INCOMPLETE DATA AND MANY TREATMENTS , 2006, Econometric Theory.

[24]  L. J. Savage,et al.  The Foundations of Statistics , 1955 .

[25]  G. Parmigiani Minimax, information and ultrapessimism , 1992 .

[26]  Jörg Stoye Minimax Regret Treatment Choice with Finite Samples and Missing Outcome Data , 2007 .

[27]  Leonard J. Savage,et al.  The Theory of Statistical Decision , 1951 .

[28]  Joshua D. Angrist,et al.  Identification of Causal Effects Using Instrumental Variables , 1993 .

[29]  D. Bergemann,et al.  Pricing Without Priors , 2007 .

[30]  Gary Chamberlain,et al.  Econometrics and decision theory , 2000 .

[31]  K. Arrow,et al.  The New Palgrave Dictionary of Economics , 2020 .

[32]  K. Hirano,et al.  Asymptotics for Statistical Treatment Rules , 2009 .

[33]  L. J. Savage,et al.  The nonexistence of certain statistical procedures in nonparametric problems , 1956 .

[34]  Paul L. Canner,et al.  Selecting One of Two Treatments When the Responses are Dichotomous , 1970 .

[35]  C. Manski Partial Identification of Probability Distributions , 2003 .

[36]  Charles F. Manski,et al.  DIVERSIFIED TREATMENT UNDER AMBIGUITY , 2008 .