Least squares is probably the best known method for fitting linear models and by far the most widely used. Surprisingly, the discrete L 1 analogue, least absolute deviations (LAD) seems to have been considered first. Possibly the LAD criterion was forced into the background because of the com putational difficulties associated with it. Recently there has been a resurgence of interest in LAD. It was spurred on by work that has resulted in efficient al gorithms for obtaining LAD fits. Another stimulus came from robust statistics. LAD estimates resist undue effects from a feyv, large errors. Therefore. in addition to being robust, they also make good starting points for other iterative, robust procedures. The LAD criterion has great utility. LAD fits are optimal for linear regressions where the errors are double exponential. However they also have excellent properties well outside this narrow context. In addition they are useful in other linear situations such as time series and multivariate data analysis. Finally, LAD fitting embodies a set of ideas that is important in linear optimization theory and numerical analysis. viii PREFACE In this monograph we will present a unified treatment of the role of LAD techniques in several domains. Some of the material has appeared in recent journal papers and some of it is new. This presentation is organized in the following way. There are three parts, one for Theory, one for Applicatior.s and one for Algorithms.

Most biologists use nonlinear regression more than any other statistical technique, but there are very few places to learn about curve-fitting. This book, by the author of the very successful Intuitive Biostatistics, addresses this relatively focused need of an extraordinarily broad range of scientists.

The least absolute deviations criterion, or the l1 norm, is frequently used for approximation where the data may contain outliers or wild points'. One of the most popular methods for solving the least absolute deviations data fitting problem is the Barrodale and Roberts (BR) algorithm (1973), which is based on linear programming techniques and the use of a modified simplex method. This algorithm is particularly efficient. However, since it is based upon the simplex method it can be susceptible to the accumulation of unrecoverable rounding errors caused by using an inappropriate pivot. In this paper we shall show how we can extend a numerically stable form of the simplex method to the special case of l1 approximation whilst still maintaining the efficiency of the Barrodale and Roberts algorithm. This extension is achieved by using the l1 characterization to rebuild the relevant parts of the simplex tableau at each iteration. The advantage of this approach is demonstrated most effectively when the observation matrix of the approximation problem is sparse, as in the case when using compactly supported basis functions such as B-splines. Under these circumstances the new method is considerably more efficient than the Barrodale and Roberts algorithm as well as being more robust.

Complete with valuable FORTRAN programs that help solve nondifferentiable nonlinear LtandLo.-norm estimation problems, this important reference/text extensively delineates ahistory of Lp-norm estimation. It examines the nonlinear Lp-norm estimation problem that isa viable alternative to least squares estimation problems where the underlying errordistribution is nonnormal, i.e., non-Gaussian.Nonlinear LrNorm Estimation addresses both computational and statistical aspects ofLp-norm estimation problems to bridge the gap between these two fields . . . contains 70useful illustrations ... discusses linear Lp-norm as well as nonlinear Lt, Lo., and Lp-normestimation problems . . . provides all appropriate computational algorithms and FORTRANlistings for nonlinear Lt- and Lo.-norm estimation problems . . . guides readers with clear endof-chapter notes on related topics and outstanding research publications . . . contains numericalexamples plus several practical problems .. . and shows how the data can prescribe variousapplications of Lp-norm alternatives.Nonlinear Lp-Norm Estimation is an indispensable reference for statisticians,operations researchers, numerical analysts, applied mathematicians, biometricians, andcomputer scientists, as well as a text for graduate students in statistics or computer science.

Design of Experiments in Nonlinear Models: Asymptotic Normality, Optimality Criteria and Small-Sample Properties provides a comprehensive coverage of the various aspects of experimental design for nonlinear models. The book contains original contributions to the theory of optimal experiments that will interest students and researchers in the field. Practitionners motivated by applications will find valuable tools to help them designing their experiments. The first three chapters expose the connections between the asymptotic properties of estimators in parametric models and experimental design, with more emphasis than usual on some particular aspects like the estimation of a nonlinear function of the model parameters, models with heteroscedastic errors, etc. Classical optimality criteria based on those asymptotic properties are then presented thoroughly in a special chapter. Three chapters are dedicated to specific issues raised by nonlinear models. The construction of design criteria derived from non-asymptotic considerations (small-sample situation) is detailed. The connection between design and identifiability/estimability issues is investigated. Several approaches are presented to face the problem caused by the dependence of an optimal design on the value of the parameters to be estimated. A survey of algorithmic methods for the construction of optimal designs is provided.