Practical Point Estimation from Higher-Order Pivots

Point estimators for a scalar parameter of interest in the presence of nuisance parameters can be defined as zero-level confidence intervals as explained in Skovgaard (1989). A natural implementation of this approach is based on estimating equations obtained from higher-order pivots for the parameter of interest. In this paper, generalising the results in Pace and Salvan (1999) outside exponential families, we take as an estimating function the modified directed likelihood. This is a higher-order pivotal quantity that can be easily computed in practice for a wide range of models, using recent advances in higher-order asymptotics (HOA, 2000). The estimators obtained from these estimating equations are a refinement of the maximum likelihood estimators, improving their small sample properties and keeping equivariance under reparameterisation. Simple explicit approximate versions of these estimators are also derived and have the form of the maximum likelihood estimator plus a function of derivatives of the loglikelihood function. Some examples and simulation studies are discussed for widely-used model classes.

[1]  I. Skovgaard,et al.  An explicit large-deviation approximation to one-parameter tests , 1996 .

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

[3]  Thomas A. Severini,et al.  An approximation to the modified profile likelihood function , 1998 .

[4]  Donald A. Pierce,et al.  Practical Use of Higher Order Asymptotics for Multiparameter Exponential Families , 1992 .

[5]  T. Severini.  Likelihood Methods in Statistics , 2001 .

[6]  K. Liang.  Estimating functions and approximate conditional likelihood , 1987 .

[7]  D. Cox,et al.  Parameter Orthogonality and Approximate Conditional Inference , 1987 .

[8]  Mik Wisniewski,et al.  Applied Regression Analysis: A Research Tool , 1990 .

[9]  O. Barndorff-Nielsen.  Modified signed log likelihood ratio , 1991 .

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

[11]  D. Cox,et al.  Inference and Asymptotics , 1994 .

[12]  A. Brazzale.  Practical small-sample parametric inference , 2000 .

[13]  Thomas A. Severini,et al.  An empirical adjustment to the likelihood ratio statistic , 1999 .

[14]  M Davidian,et al.  The effect of variance function estimation on nonlinear calibration inference in immunoassay data. , 1996, Biometrics.

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

[16]  Luigi Pace,et al.  The directed modified profile likelihood in models with many nuisance parameters , 1999 .

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

[18]  J. O. Rawlings,et al.  Applied Regression Analysis: A Research Tool , 1988 .

[19]  I. Skovgaard.  A General Large Deviation Approximation to One-parameter Tests a General Large Deviation Approximation to One-parameter Tests , 1995 .

[20]  O. Barndorff-Nielsen.  A note on the standardized signed log likelihood ratio , 1990 .