Median bias reduction of maximum likelihood estimates

For regular parametric problems, we show how median centering of the maximum likelihood estimate can be achieved by a simple modification of the score equation. For a scalar parameter of interest, the estimator is equivariant under interest respecting parameterizations and third-order median unbiased. With a vector parameter of interest, componentwise equivariance and third-order median centering are obtained. Like Firth's (1993, Biometrika) implicit method for bias reduction, the new method does not require finiteness of the maximum likelihood estimate and is effective in preventing infinite estimates. Simulation results for continuous and discrete models, including binary and beta regression, confirm that the method succeeds in achieving componentwise median centering and in solving the infinite estimate problem, while keeping comparable dispersion and the same approximate distribution as its main competitors.

[1]  Ioannis Kosmidis,et al.  Bias in parametric estimation: reduction and useful side‐effects , 2013, 1311.6311.

[2]  M. Schemper,et al.  A solution to the problem of separation in logistic regression , 2002, Statistics in medicine.

[3]  H. Holling,et al.  Saddlepoint Approximations of the Distribution of the Person Parameter in the Two Parameter Logistic Model , 2015, Psychometrika.

[4]  D. Cox,et al.  Asymptotic techniques for use in statistics , 1989 .

[5]  Hsiuying Wang,et al.  TOLERANCE INTERVALS FOR DISCRETE DISTRIBUTIONS IN EXPONENTIAL FAMILIES , 2009 .

[6]  Robert Tibshirani,et al.  A Simple Method for the Adjustment of Profile Likelihoods , 1990 .

[7]  Debashis Kushary,et al.  Bootstrap Methods and Their Application , 2000, Technometrics.

[8]  Alessandra Salvan,et al.  Point estimation based on confidence intervals: exponential families , 1999 .

[9]  David Firth,et al.  Bias reduction in exponential family nonlinear models , 2009 .

[10]  O. Barndorff-Nielsen.  On a formula for the distribution of the maximum likelihood estimator , 1983 .

[11]  William E. Griffiths,et al.  Learning and Practicing Econometrics , 1993 .

[12]  P. Hall.  The Bootstrap and Edgeworth Expansion , 1992 .

[13]  O. E. Barndorff-Nielsen,et al.  Infereni on full or partial parameters based on the standardized signed log likelihood ratio , 1986 .

[14]  Laura Ventura,et al.  Practical Point Estimation from Higher-Order Pivots , 2002 .

[15]  Thomas J. DiCiccio,et al.  Information Bias and Adjusted Profile Likelihoods , 1996 .

[16]  Achim Zeileis,et al.  Extended Beta Regression in R: Shaken, Stirred, Mixed, and Partitioned , 2012 .

[17]  G. Alastair Young.  ROUTES TO HIGHER-ORDER ACCURACY IN PARAMETRIC INFERENCE , 2009 .

[18]  Luigi Pace,et al.  Principles of statistical inference : from a neo-Fisherian perspective , 1997 .

[19]  N. Sartori.  Bias prevention of maximum likelihood estimates for scalar skew normal and skew t distributions , 2006 .

[20]  A. Agresti.  Foundations of Linear and Generalized Linear Models , 2015 .

[21]  D. Firth.  Bias reduction of maximum likelihood estimates , 1993 .

[22]  Nicola Sartori,et al.  Modified profile likelihoods in models with stratum nuisance parameters , 2003 .

[23]  M. N. Vrahatis,et al.  From linear to nonlinear iterative methods , 2003 .

[24]  Norman E. Breslow,et al.  Odds ratio estimators when the data are sparse , 1981 .

[25]  Anastasios A. Tsiatis,et al.  Median Unbiased Estimation for Binary Data , 1989 .

[26]  Ioannis Kosmidis,et al.  A generic algorithm for reducing bias in parametric estimation , 2011 .

[27]  Bent Jørgensen,et al.  Parameter Orthogonality and Bias Adjustment for Estimating Functions , 2004 .

[28]  A note on conditional cumulants in canonical exponential families , 1992 .

[29]  C. Read.  Median Unbiased Estimators , 2014 .

[30]  Steven Stern,et al.  A Second‐order Adjustment to the Profile Likelihood in the Case of a Multidimensional Parameter of Interest , 1997 .

[31]  E. Lehmann.  Testing Statistical Hypotheses , 1960 .