Valid post-selection inference

It is common practice in statistical data analysis to perform data-driven variable selection and derive statistical inference from the resulting model. Such inference enjoys none of the guarantees that classical statistical theory provides for tests and confidence intervals when the model has been chosen a priori. We propose to produce valid ``post-selection inference'' by reducing the problem to one of simultaneous inference and hence suitably widening conventional confidence and retention intervals. Simultaneity is required for all linear functions that arise as coefficient estimates in all submodels. By purchasing ``simultaneity insurance'' for all possible submodels, the resulting post-selection inference is rendered universally valid under all possible model selection procedures. This inference is therefore generally conservative for particular selection procedures, but it is always less conservative than full Scheffe protection. Importantly it does not depend on the truth of the selected submodel, and hence it produces valid inference even in wrong models. We describe the structure of the simultaneous inference problem and give some asymptotic results.

[1]  R. A. Fisher,et al.  Design of Experiments , 1936 .

[2]  H. O. Lancaster,et al.  The derivation and partition of chi2 in certain discrete distributions. , 1949, Biometrika.

[3]  H. Scheffé A METHOD FOR JUDGING ALL CONTRASTS IN THE ANALYSIS OF VARIANCE , 1953 .

[4]  D. Campbell Factors relevant to the validity of experiments in social settings. , 1957, Psychological bulletin.

[5]  T D WILSON,et al.  Smoking and Lung Cancer , 1960, Journal of the Irish Medical Association.

[6]  R. Buehler,et al.  Note on a Conditional Property of Student's $t^1$ , 1963 .

[7]  R. R. Bahadur A Note on Quantiles in Large Samples , 1966 .

[8]  L. Brown THE CONDITIONAL LEVEL OF STUDENT'S t TEST' , 1967 .

[9]  A. Wyner Random packings and coverings of the unit n-sphere , 1967 .

[10]  D. Campbell Reforms as experiments , 1969 .

[11]  J. Gart An exact test for comparing matched proportions in crossover designs , 1969 .

[12]  R. Olshen The Conditional Level of the F—Test , 1973 .

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

[14]  D. Cox A note on data-splitting for the evaluation of significance levels , 1975 .

[15]  John W. Tukey,et al.  Data Analysis and Regression: A Second Course in Statistics , 1977 .

[16]  P. Sen Asymptotic Properties of Maximum Likelihood Estimators Based on Conditional Specification , 1979 .

[17]  D. Rubin,et al.  Assessing Sensitivity to an Unobserved Binary Covariate in an Observational Study with Binary Outcome , 1983 .

[18]  R. Freeman Longitudinal Analysis of the Effect of Trade Unions , 1984 .

[19]  Jerry A. Hausman,et al.  Errors in Variables in Panel Data , 1984 .

[20]  Richard B. Freeman,et al.  Longitudinal Analyses of the Effects of Trade Unions , 1983, Journal of Labor Economics.

[21]  K E Warner,et al.  Smoking and lung cancer: an overview. , 1984, Cancer research.

[22]  P. Rosenbaum From Association to Causation in Observational Studies: The Role of Tests of Strongly Ignorable Treatment Assignment , 1984 .

[23]  Takashi Yanagawa,et al.  Case-control studies: Assessing the effect of a confounding factor , 1984 .

[24]  William M. K. Trochim,et al.  Pattern Matching, Validity, and Conceptualization in Program Evaluation , 1985 .

[25]  D. Rubin,et al.  Constructing a Control Group Using Multivariate Matched Sampling Methods That Incorporate the Propensity Score , 1985 .

[26]  P. Sen,et al.  On preliminary test and shrinkage m-estimation in linear models , 1987 .

[27]  Raymond J. Carroll,et al.  Variance Function Estimation in Regression: the Effect of Estimating the Mean , 1988 .

[28]  Theo K. Dijkstra,et al.  Data-driven selection of regressors and the bootstrap , 1988 .

[29]  Paul R. Rosenbaum,et al.  Optimal Matching for Observational Studies , 1989 .

[30]  Clifford M. Hurvich,et al.  The impact of model selection on inference in linear regression , 1990 .

[31]  L. Brown An Ancillarity Paradox Which Appears in Multiple Linear Regression , 1990 .

[32]  H. J. Arnold Introduction to the Practice of Statistics , 1990 .

[33]  D. Rubin [On the Application of Probability Theory to Agricultural Experiments. Essay on Principles. Section 9.] Comment: Neyman (1923) and Causal Inference in Experiments and Observational Studies , 1990 .

[34]  B. M. Pötscher Effects of Model Selection on Inference , 1991, Econometric Theory.

[35]  E. Mammen Bootstrap and Wild Bootstrap for High Dimensional Linear Models , 1993 .

[36]  E. Ziegel Introduction to the Practice of Statistics (2nd ed.) , 1994 .

[37]  Paul R. Rosenbaum,et al.  Quantiles in Nonrandom Samples and Observational Studies , 1995 .

[38]  S. Marcus,et al.  Using Omitted Variable Bias to Assess Uncertainty in the Estimation of an AIDS Education Treatment Effect , 1997 .

[39]  R. Kronmal,et al.  Assessing the sensitivity of regression results to unmeasured confounders in observational studies. , 1998, Biometrics.

[40]  Paul Kabaila,et al.  VALID CONFIDENCE INTERVALS IN REGRESSION AFTER VARIABLE SELECTION , 1998, Econometric Theory.

[41]  P. Rosenbaum,et al.  Dual and simultaneous sensitivity analysis for matched pairs , 1998 .

[42]  J. Angrist,et al.  Empirical Strategies in Labor Economics , 1998 .

[43]  J. Robins,et al.  Sensitivity Analysis for Selection bias and unmeasured Confounding in missing Data and Causal inference models , 2000 .

[44]  M. Bhaskara Rao,et al.  Model Selection and Inference , 2000, Technometrics.

[45]  Hannes Leeb,et al.  The Finite-Sample Distribution of Post-Model-Selection Estimators, and Uniform Versus Non-Uniform Approximations , 2000 .

[46]  Robert Tibshirani,et al.  The Elements of Statistical Learning , 2001 .

[47]  W. Shadish,et al.  Experimental and Quasi-Experimental Designs for Generalized Causal Inference , 2001 .

[48]  J. Copas,et al.  Local sensitivity approximations for selectivity bias , 2001 .

[49]  Alberto Abadie,et al.  The Economic Costs of Conflict: A Case Study of the , 2003 .

[50]  Hannes Leeb,et al.  Performance Limits for Estimators of the Risk or Distribution of Shrinkage-Type Estimators, and Some General Lower Risk-Bound Results , 2002 .

[51]  N. Hjort,et al.  Frequentist Model Average Estimators , 2003 .

[52]  M.,et al.  THE FINITE-SAMPLE DISTRIBUTION OF POST-MODEL-SELECTION ESTIMATORS AND UNIFORM VERSUS NONUNIFORM APPROXIMATIONS , 2003, Econometric Theory.

[53]  H. Leeb,et al.  CAN ONE ESTIMATE THE UNCONDITIONAL DISTRIBUTION OF POST-MODEL-SELECTION ESTIMATORS? , 2003, Econometric Theory.

[54]  N. Hjort,et al.  The Focused Information Criterion , 2003 .

[55]  G. W. Imbens Sensitivity to Exogeneity Assumptions in Program Evaluation , 2003 .

[56]  Paul R. Rosenbaum,et al.  Design sensitivity in observational studies , 2004 .

[57]  T. DiPrete,et al.  7. Assessing Bias in the Estimation of Causal Effects: Rosenbaum Bounds on Matching Estimators and Instrumental Variables Estimation with Imperfect Instruments , 2004 .

[58]  D. Ruppert The Elements of Statistical Learning: Data Mining, Inference, and Prediction , 2004 .

[59]  I. Verdinelli,et al.  False Discovery Control for Random Fields , 2004 .

[60]  B. M. Pötscher,et al.  MODEL SELECTION AND INFERENCE: FACTS AND FICTION , 2005, Econometric Theory.

[61]  Hannes Leeb,et al.  The distribution of a linear predictor after model selection: Unconditional finite-sample distributions and asymptotic approximations , 2005, math/0611186.

[62]  J. Ioannidis Why Most Published Research Findings Are False , 2005, PLoS medicine.

[63]  J. Ioannidis Why Most Published Research Findings Are False , 2005 .

[64]  B. Hansen,et al.  Optimal Full Matching and Related Designs via Network Flows , 2006 .

[65]  A. K. Md. Ehsanes Saleh,et al.  Theory of preliminary test and Stein-type estimation with applications , 2006 .

[66]  Changes in Service Availability in California Hospitals, 1995 to 2002 , 2006, Journal of healthcare management / American College of Healthcare Executives.

[67]  B. M. Pötscher The Distribution of Model Averaging Estimators and an Impossibility Result Regarding Its Estimation , 2006 .

[68]  Paul Kabaila,et al.  On the Large-Sample Minimal Coverage Probability of Confidence Intervals After Model Selection , 2006 .

[69]  The distribution of a linear predictor after model selection: Unconditional finite-sample distributions and asymptotic approximations , 2006, math/0611186.

[70]  H. Leeb,et al.  Sparse Estimators and the Oracle Property, or the Return of Hodges' Estimator , 2007, 0704.1466.

[71]  Benedikt M. Potscher,et al.  On the distribution of the adaptive LASSO estimator , 2008, 0801.4627.

[72]  Theory of Preliminary Test and Stein-Type Estimation With Applications , 2007 .

[73]  G. Imbens,et al.  Bias-Corrected Matching Estimators for Average Treatment Effects , 2002 .

[74]  B. M. Pötscher,et al.  CAN ONE ESTIMATE THE UNCONDITIONAL DISTRIBUTION OF POST-MODEL-SELECTION ESTIMATORS? , 2007, Econometric Theory.

[75]  Benedikt M. Potscher,et al.  Confidence sets based on penalized maximum likelihood estimators in Gaussian regression , 2008, 0806.1652.

[76]  David R. Holtgrave,et al.  Alternatives to the randomized controlled trial. , 2008, American journal of public health.

[77]  Joshua D. Angrist,et al.  Mostly Harmless Econometrics: An Empiricist's Companion , 2008 .

[78]  Stephen W Lagakos,et al.  Inference after variable selection using restricted permutation methods. , 2009, The Canadian journal of statistics = Revue canadienne de statistique.

[79]  Dylan S. Small,et al.  Split Samples and Design Sensitivity in Observational Studies , 2009 .

[80]  Paul Kabaila The Coverage Properties of Confidence Regions After Model Selection , 2009 .

[81]  P. Rosenbaum Design of Observational Studies , 2009, Springer Series in Statistics.

[82]  Benedikt M. Pötscher,et al.  On the Distribution of Penalized Maximum Likelihood Estimators: The LASSO, SCAD, and Thresholding , 2007, J. Multivar. Anal..

[83]  Robert Tibshirani,et al.  The Elements of Statistical Learning: Data Mining, Inference, and Prediction, 2nd Edition , 2001, Springer Series in Statistics.

[84]  Paul R. Rosenbaum,et al.  Sensitivity Analysis for Equivalence and Difference in an Observational Study of Neonatal Intensive Care Units , 2009 .

[85]  Richard A. Berk,et al.  Statistical Inference After Model Selection , 2010 .

[86]  B. Poindexter,et al.  Neonatal Outcomes of Extremely Preterm Infants From the NICHD Neonatal Research Network , 2010, Pediatrics.

[87]  P. Rosenbaum Evidence factors in observational studies , 2010 .

[88]  Elizabeth A Stuart,et al.  Matching methods for causal inference: A review and a look forward. , 2010, Statistical science : a review journal of the Institute of Mathematical Statistics.

[89]  Paul R. Rosenbaum,et al.  Design Sensitivity and Efficiency in Observational Studies , 2010 .

[90]  Ulrike Schneider,et al.  Distributional results for thresholding estimators in high-dimensional Gaussian regression models , 2011, 1106.6002.

[91]  Paul R. Rosenbaum,et al.  Some Approximate Evidence Factors in Observational Studies , 2011 .

[92]  L. Oakley,et al.  The effectiveness of antenatal care programmes to reduce infant mortality and preterm birth in socially disadvantaged and vulnerable women in high-income countries: a systematic review , 2011, BMC pregnancy and childbirth.

[93]  S. Stanley Young,et al.  Deming, data and observational studies , 2011 .

[94]  Dylan S. Small,et al.  Using Split Samples and Evidence Factors in an Observational Study of Neonatal Outcomes , 2011 .

[95]  Walter Zucchini,et al.  Model Selection , 2011, International Encyclopedia of Statistical Science.

[96]  Rembert De Blander,et al.  Mostly Harmless Econometrics: An Empiricist's Companion , 2011 .

[97]  A. Young Mostly Harmless Econometrics , 2012 .