## Configural polysampling: a route to practical robustnessProvides a broad readership--both novice and advanced students and researchers--with a solid introduction to conditional thinking in a robust setting. Using a minimum of mathematics, it considers three types of parameters: location, scale and regression slopes. Recognizing their importance to performance evaluation, the book focuses on confrontations of a small number of suitably disparate shapes as well as on invariance. |

### What people are saying - Write a review

We haven't found any reviews in the usual places.

### Contents

Background | 1 |

IF Summary | 7 |

2C A Brief History of the Last 50 Years of the Location | 13 |

Copyright | |

25 other sections not shown

### Common terms and phrases

1991 John Wiley analysis ancillary statistic apply asymptotically equivalent average behavior biconditional bieffective estimator biefficient biefficient-like CMLE bioptimal estimator bioptimal intervals calculate centercept Chapter cMSEF(T compute conditional coverage conditional density conditional mean-squared error confidence interval configural estimators configural polysampling corresponding coverage probability defined denotes derived discussed distribution F distributional shapes efficiency equal equivariant estimator estimator of location evaluate example EXHIBIT formula Gaussian distribution Gaussian rule Gaussian situation Huber idea influence function integral integrand kurtosis Lagrange multipliers Laplace approximation Laplace approximation method least-squares linear location estimator M-estimators mators maximum likelihood estimators median minimizes neighborhood numerical optimal compromise estimators performance Pitman estimator plot Polyefficiency problem pushback relative excess variances robust estimators Robust Statistics sample size sample sizes sampling situations Section shadow prices shows simple slash distributions stochastic Student's distributions symmetric techniques tion transformation Trimmed mean Tukey tuning constant tuning values underlying distribution vector weights zero