Sampling

This book is an interesting contribution to the literature on sampling of finite populations and other related material on applied sampling such as methods for elusive and hard-to-detect populations, spatial sampling and adaptive sampling. Two articles by the reviewer [1,2] are similar in content to the new sections of the third edition of the book. This material does not appear in the first two editions of the same book, which demonstrates that the book is updated in several aspects. Part of the new content is not proven in the book and no reference is given by the author to justify the results. In the formula of the variance estimation for the nonresponse estimator of the population mean, I would say that the number of strata is two, and then it is better to substitute L for 2 in the notation. The general formulae (14.1) and (14.2) for stratified double sampling or postgrouped sampling are valid for any natural value of L, but (14.3) and the respective variance estimation must be simplified to 2 strata. I think that the content of the book could be simplified by the treatment of simple random sampling with replacement as the basic method of sampling. It is traditional to present simple random sampling without replacement as more efficient than simple random sampling with replacement, but for a common fixed sample size, the expected cost for this is lesser than for that with supposing only a fixed cost per observed unit. Keeping memory of the sampled units, the multiplicity in the sample of the same unit is computed only with the same cost as one effective sampled unit, and hence it reduces its expected cost over simple random sampling without replacement with equal sample size where the multiplicity of a population unit is not possible. I highly recommend this book over many statistical ones in which the interest is put on subjective inferences as, for example, in classical parametric, Bayesian, nonparametric, and superpopulation inferences. But my recommendation is not total because this edition does not offer proofs of some new results.