Statistical inference

"In many ways, estimation by an appropriate minimum distance method is one of the most natural ideas in statistics. However, there are many different ways of constructing an appropriate distance between the data and the model: the scope of study referred to by "Minimum Distance Estimation" is literally huge. Filling a statistical resource gap, Statistical Inference: The Minimum Distance Approach comprehensively overviews developments in density-based minimum distance inference for independently and identically distributed data. Extensions to other more complex models are also discussed. Comprehensively covering the basics and applications of minimum distance inference, this book introduces and discusses: The estimation and hypothesis testing problems for both discrete and continuous modelsThe robustness properties and the structural geometry of the minimum distance methodsThe inlier problem and its possible solutions, and the weighted likelihood estimation problem The extension of the minimum distance methodology in interdisciplinary areas, such as neural networks and fuzzy sets, as well as specialized models and problems, including semi-parametric problems, mixture models, grouped data problems and survival analysis. Statistical Inference: The Minimum Distance Approach gives a thorough account of density-based minimum distance methods and their use in statistical inference. It covers statistical distances, density-based minimum distance methods, discrete and continuous models, asymptotic distributions, robustness, computational issues, residual adjustment functions, graphical descriptions of robustness, penalized and combined distances, weighted likelihood, and multinomial goodness-of-fit tests. This carefully crafted resource is useful to researchers and scientists within and outside the statistics arena"-- "Preface In many ways, estimation by an appropriate minimum distance method is one of the most natural ideas in statistics. A parametric model imposes a certain structure on the class of probability distributions that may be used to describe real life data generated from a process under study. There hardly appears to be a better way to deal with such a problem than to choose the parametric model that minimizes an appropriately defined distance between the data and the model. The issue is an important and complex one. There are many different ways of constructing an appropriate "distance" between the "data" and the "model". One could, for example, construct a distance between the empirical distribution function and the model distribution function by a suitable measure of distance. Alternatively, one could minimize the distance between the estimated data density (obtained, if necessary, by using a nonparametric smoothing technique such as kernel density estimation) and the parametric model density. And when the particular nature of the distances have been settled (based on distribution functions, based on densities, etc.), there may be innumerable options for the distance to be used within the particular type of distances. So the scope of study referred to by "Minimum Distance Estimation" is literally huge"--