Sampling of Populations

Methods and Applications

Paul S. Levy, Stanley Lemeshow

• Publisher: Wiley
• Year: 2013
• ISBN:
  9781118627310
• Edition: 4
• Topic: Medicine
• Subtopic: Biostatistics

A trusted classic on the key methods in population sampling—now in a modernized and expanded new edition

Sampling of Populations, Fourth Edition continues to serve as an all-inclusive resource on the basic and most current practices in population sampling. Maintaining the clear and accessible style of the previous edition, this book outlines the essential statistical methodsfor survey design and analysis, while also exploring techniques that have developed over the past decade.

The Fourth Edition successfully guides the reader through the basic concepts and procedures that accompany real-world sample surveys, such as sampling designs, problems of missing data, statistical analysis of multistage sampling data, and nonresponse and poststratification adjustment procedures. Rather than employ a heavily mathematical approach, the authors present illustrative examples that demonstrate the rationale behind common steps in the sampling process, from creating effective surveys to analyzing collected data. Along with established methods, modern topics are treated through the book's new features, which include:

• A new chapter on telephone sampling, with coverage of declining response rates, the creation of "do not call" lists, and the growing use of cellular phones
• A new chapter on sample weighting that focuses on adjustments to weight for nonresponse, frame deficiencies, and the effects of estimator instability
• An updated discussion of sample survey data analysis that includes analytic procedures for estimation and hypothesis testing
• A new section on Chromy's widely used method of taking probability proportional to size samples with minimum replacement of primary sampling units
• An expanded index with references on the latest research in the field

All of the book's examples and exercises can be easily worked out using various software packages including SAS, STATA, and SUDAAN, and an extensive FTP site contains additional data sets. With its comprehensive presentation and wealth of relevant examples, Sampling of Populations, Fourth Edition is an ideal book for courses on survey sampling at the upper-undergraduate and graduate levels. It is also a valuable reference for practicing statisticians who would like to refresh their knowledge of sampling techniques.

PART 1

Basic Concepts

CHAPTER 1

Uses of Sample Surveys

1.1 WHY SAMPLE SURVEYS ARE USED

Information on characteristics of populations is constantly needed by politicians, marketing departments of companies, public officials responsible for planning health and social services, and others. For reasons relating to timeliness and cost, this information is often obtained by use of sample surveys. Such surveys are the subject of this book.

The following discussion provides an example of a sample survey conducted to obtain information about a health characteristic in a particular population. A health department in a large state is interested in determining the proportion of the state’s children of elementary school age who have been immunized against childhood infectious diseases (e.g., polio, diphtheria, tetanus, and pertussis). For administrative reasons, this task must be completed in only one month.

At first glance this task would seem to be most formidable, involving the careful coordination of a large staff attempting to collect information, either from parents or from school immunization records on each and every child of elementary school age residing in that state. Clearly, the budget necessary for such an undertaking would be enormous because of the time, travel expenses, and number of children involved. Even with a sizable staff, it would be difficult to complete such an undertaking in the specified time frame.

To handle problems such as the one outlined above, this text will present a variety of methods for selecting a subset (a sample) from the original set of all measurements (the population) of interest to the researchers. It is the members of the sample who will be interviewed, studied, or measured. For example, in the problem stated above, the net effect of such methods will be that valid and reliable estimates of the proportion of children who have been immunized for these diseases could be obtained in the time frame specified and at a fraction of the cost that would have resulted if attempts were made to obtain the information concerning every child of elementary school age in the state.

More formally, a sample survey may be defined as a study involving a subset (or sample) of individuals selected from a larger population. Variables or characteristics of interest are observed or measured on each of the sampled individuals. These measurements are then aggregated over all individuals in the sample to obtain summary statistics (e.g., means, proportions, and totals) for the sample. It is from these summary statistics that extrapolations can be made concerning the entire population. The validity and reliability of these extrapolations depend on how well the sample was chosen and on how well the measurements were made. These issues constitute the subject matter of this text.

When all individuals in the population are selected for measurement, the study is called a census. The summary statistics obtained from a census are not extrapolations, since every member of the population is measured. The validity of the resulting statistics, however, depends on how well the measurements are made. The main advantages of sample surveys over censuses lie in the reduced costs and greater speed made possible by taking measurements on a subset rather than on an entire population. In addition, studies involving complex issues requiring elaborate measurement procedures are often feasible only if a sample of the population is selected for measurement since limited resources can be allocated to getting detailed measurements if the number of individuals to be measured is not too great.

In the United States, as in many other countries, governmental agencies are mandated to develop and maintain programs whereby sample surveys are used to collect data on the economic, social, and health status of the population, and these data are used for research purposes as well as for policy decisions. For example, the National Center for Health Statistics (NCHS), a center within the United States Department of Health and Human Services, is mandated by law to conduct a program of periodic and ongoing sample surveys designed to obtain information about illness, disability, and the utilization of health care services in the United States [15]. Similar agencies, centers, or bureaus exist within other departments (e.g., the Bureau of Labor Statistics within the Department of Labor, and the National Center for Educational Statistics within the Department of Education) that collect data relevant to the mission of their departments through a program of sample surveys. Field work for these surveys is sometimes done by the U.S. Bureau of the Census, which also has its own program of surveys, or by commercial firms.

The surveys developed by such government agencies often have extremely complex designs and require very large and highly skilled staff (and, hence, large budgets) for their execution. Although the nature of the missions of these government agencies—provision of valid and reliable statistics on a wide variety of indicators for the United States as a whole and various subgroups of it—would justify these large budgets, such costs are rarely justified or at all feasible for most institutions that make use of sample surveys. The information needs of most potential users of sample surveys are far more limited in scope and much more focused around a relatively small set of particular questions. Thus, the types of surveys conducted outside of the federal government are generally simpler in design and “one-shot” rather than ongoing. These are the types of surveys on which we will focus in this text. We will, however, devote some discussion to more complex sample surveys, especially in Chapter 12, which discusses variance estimation methods that have been developed primarily to meet the needs of very complex government surveys.

Sample surveys belong to a larger class of nonexperimental studies generally given the name “observational studies” in the health or social sciences literature. Most sample surveys can be put in the class of observational studies known as “cross-sectional studies.” Other types of observational studies include cohort studies and case-control studies.

Cross-sectional studies are “snapshots” of a population at a single point in time, having as objectives either the estimation of the prevalence or the mean level of some characteristics of the population or the measurement of the relationship between two or more variables measured at the same point in time.

Cohort and case-control studies are used for analytic rather than for descriptive purposes. For example, they are used in epidemiology to test hypotheses about the association between exposure to suspected risk factors and the incidence of specific diseases.

These study designs are widely used to gain insight into relationships. In the business world, for example, a sample of delinquent accounts might be taken (i.e., the “cases”) along with a sample of accounts that are not delinquent (i.e., the “controls”), and the characteristics of each group might be compared for purposes of determining those factors that are associated with delinquency. Numerous examples of these study designs could be given in other fields;

As mentioned above, cohort and case-control studies are designed with the objec...