Indirect Sampling: (Springer Series in Statistics)

Following the classical sampling theory, the survey statistician selects samples of people, businesses or others, in order to obtain the desired information. Drawing the samples is usually done by randomly selecting from a list representing the target population. In practice, this list is often not available. At best, the statistician only has access to a different list, indirectly related to the targeted population. The example of a survey of children where the statistician only has a list of adult persons is a typical case. In this case, the statistician first draws a sample of adults, and for each selected adult, the statistician then identifies his/her children. The survey is done from the latter. This is what is called indirect sampling. When indirect sampling is used jointly with the sampling of clusters of persons (families, for example), many complications arise for the survey statistician. One of the complications relates to the computation of the estimates from the survey. The production of estimates of simple totals or means can then become nightmares for the survey statistician. To solve this problem, the author proposes a simple solution, easy to implement, that is called the generalised weight share method. This book is the reference on indirect sampling and the generalised weight share method. It contains the different developments done by the author on these subjects. The theory surrounding them is presented, but also different possible applications that drive its interest. The reader will find in this book the answer to questions that come, inevitably, when working in a context of indirect sampling.

Table of Contents:
Description and Use of the GWSM.- Literature Review.- Properties.- Other Generalisations.- Application in Longitudinal Surveys.- GWSM and Calibration.- Non-response.- GWSM and Record Linkage.- Conclusion.

Review :
From the reviews: "The author of the book wrote a doctoral thesis on indirect sampling, and this book evolved around his doctoral work. He adopts the classical design-based point of view: inferences are principally made with respect to the sampling plan. In general, the book is written with care and it represents a welcome contribution to the survey literature. … It is likely that the audience of the book will be mainly composed of survey statisticians, researchers in sampling theory and graduate students." (Pierre Duchesne, Mathematical Reviews, Issue 2009 e) “The classic paradigm for sampling inference is based on drawing a probability sample from a target population and then making an estimate of a population parameter of interest…based on the sample data. Of course, this presupposes that the target population is one from which a sample can be drawn, but in many situations this is not possible. This book is about what sometimes can be done instead, when we have access to another population that can be sampled and it is possible using information collected from this second, feasible sample to indentify some of the units in the target population, and thus to measure their values of Y. The set of values obtained in this way is usually referred to as an indirect sample from the target population. …The author has been one of the main contributors to the development of a method of sample weighting, the generalized weight share method (GWSM), that can be used with such indirect samples. This method is the focus of this book. …Overall, an experienced survey methodologist should find this book interesting…” ((Journal of the American Statistical Association, September 2009, Vol. 104, No. 487) “Indirect Sampling is a book which contains ideas and results when the target population is not available, but the statistician has access to an indirectly related different list. … the author explains a theory useful togovernments, company managers, sociologists, economists, or ordinary citizens to take a decision with the information provided by surveys when the available frame has defects in the list of units of the population. It could be used too in engineering applications … .” (Mariano Ruiz Espejo, Technometrics, Vol. 53 (3), August, 2011)