Skip to main content
Fachgebiete
Automobil + Motoren Bauwesen + Immobilien Business IT + Informatik Elektrotechnik + Elektronik Energie + Nachhaltigkeit Finance + Banking Management + Führung Marketing + Vertrieb Maschinenbau + Werkstoffe Versicherung + Risiko
DE
EN
Bücher
Zeitschriften
Themenseiten
Organisationspsychologie Projektmanagement Marketing Smart Manufacturing
Weitere Formate
Podcasts Webinare Technik Webinare Wirtschaft Kongresse Veranstaltungskalender Awards
Jetzt Einzelzugang starten
Zugang für Unternehmen
Referenzkunden
SLX-Digitalkonferenz: Zukunftswerkstatt 2024

Mehr Umsatz mit Top-Kontakten

Am 7. Mai erfahren Sie, wie Vertriebsteams Leadgenerierung, Leadnurturing und Leadstrategien effizient steuern können.
Erweiterte Suche
Anmelden
nach oben

2013 | Buch

An Overview of Sample Design and Weighting Kapitel lesen Erstes Kapitel lesen

Practical Tools for Designing and Weighting Survey Samples

verfasst von: Richard Valliant, Jill A. Dever, Frauke Kreuter

Verlag: Springer New York

Buchreihe : Statistics for Social and Behavioral Sciences

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Einloggen, um Zugang zu erhalten
insite
SUCHEN

Über dieses Buch

Survey sampling is fundamentally an applied field. The goal in this book is to put an array of tools at the fingertips of practitioners by explaining approaches long used by survey statisticians, illustrating how existing software can be used to solve survey problems, and developing some specialized software where needed. This book serves at least three audiences: (1) Students seeking a more in-depth understanding of applied sampling either through a second semester-long course or by way of a supplementary reference; (2) Survey statisticians searching for practical guidance on how to apply concepts learned in theoretical or applied sampling courses; and (3) Social scientists and other survey practitioners who desire insight into the statistical thinking and steps taken to design, select, and weight random survey samples.

Several survey data sets are used to illustrate how to design samples, to make estimates from complex surveys for use in optimizing the sample allocation, and to calculate weights. Realistic survey projects are used to demonstrate the challenges and provide a context for the solutions. The book covers several topics that either are not included or are dealt with in a limited way in other texts. These areas include: sample size computations for multistage designs; power calculations related to surveys; mathematical programming for sample allocation in a multi-criteria optimization setting; nuts and bolts of area probability sampling; multiphase designs; quality control of survey operations; and statistical software for survey sampling and estimation. An associated R package, PracTools, contains a number of specialized functions for sample size and other calculations. The data sets used in the book are also available in PracTools, so that the reader may replicate the examples or perform further analyses.

Inhaltsverzeichnis

Frontmatter
Chapter 1. An Overview of Sample Design and Weighting
Abstract
This is a practical book. Many techniques used by survey practitioners are not covered by standard textbooks but are necessary to do a professional job when designing samples and preparing data for analyses. In this book, we present a collection of methods that we have found most useful in our own practical work. Since computer software is essential in applying the techniques, example code is given throughout.
Richard Valliant, Jill A. Dever, Frauke Kreuter

Designing Single-stage Sample Surveys

Frontmatter
Chapter 2. Project 1: Design a Single-Stage Personnel Survey
Abstract
Our primary goal is to equip survey researchers with the tools needed to design and weight survey samples. This chapter gives the first of several projects that mirror some of the complexities found in applied work.
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 3. Sample Design and Sample Size for Single-Stage Surveys
Abstract
Chapter 3 covers the problem of determining a sample size for single-stage surveys with imposed constraints such as a desired level of precision. To determine a sample size, a particular type of statistic must be considered. Means, totals, and proportions are emphasized in this chapter. We concentrate on simple random samples selected without replacement in Sect.3.1. Precision targets can be set in terms of coefficients of variation or margins of error for unstratified designs as discussed in Sect. 3.1.1. We cover stratified simple random sampling in Sect. 3.1.2. Determining a sample size when sampling with varying probabilities is somewhat more complicated because the without-replacement variance formula is complex. A useful device for determining a sample size when sampling with probability proportional to size (pps) is to employ the design-based variance formula for with-replacement sampling, as covered in Sect. 3.2.1. Although we mainly cover calculations based on design-based variances, models are also especially useful when analyzing pps sampling as discussed in Sect. 3.2.2.
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 4. Power Calculations and Sample Size Determination
Abstract
In Chap. we calculated sample sizes based on targets for coefficients of variation (CV s), margins of error, and cost constraints. Another method is to determine the sample size needed to detect a particular alternative value when testing a hypothesis. For example, when comparing the means for two groups, one way of determining sample size is through a power calculation. Roughly speaking, power is a measure of how likely you are to recognize a certain size of difference in the means. A sample size is determined that will allow that difference to be detected with high probability (i.e., a detectable difference). Power can also be determined in a one-sample case where a simple hypothesis is being tested versus a simple alternative. Using power to determine sample sizes is especially useful when some important analytic comparisons can be identified in advance of selecting the sample. Although not covered in most books on sample design, most practitioners will inevitably have applications where power calculations are needed.
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 5. Mathematical Programming
Abstract
Earlier chapters examined sample size determination and allocation to strata for a single variable. In reality, almost every survey of any size is multipurpose. Data on a number of different variables are collected on each sample unit. Estimates are made of population values for the full population and for various domains or subpopulations. In addition, a variety of types of estimates may be made, including means, totals, quantiles, and model parameters.
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 6. Outcome Rates and Effect on Sample Size
Abstract
Outcome rates, such as the percent of sample units refusing to participate in a survey, generally have three uses. The first is to measure study performance and outcome rates are often also referred to as performance rates or process indicators. For example, a client might wish to know what proportion of the sample resulted in a completed interview. The second use is to inflate a calculated sample size for loss of sample units.
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 7. The Personnel Survey Design Project: One Solution
Abstract
An optimization problem was presented in Chap. 2 for a single-stage stratified sample design. In the following sections, we present a solution to the multipurpose design question borrowing from material presented in Chaps. 3–6. A series of solutions was generated for the sample allocation to test the sensitivity of the assumptions. Additionally, different software may produce different yet comparable results. Ultimately, a single solution must be chosen from this set for implementation as discussed below.
Richard Valliant, Jill A. Dever, Frauke Kreuter

Multistage Designs

Frontmatter
Chapter 8. Project 2: Designing an Area Sample
Abstract
In this project you will design a sample of census tracts, block groups, and persons from Anne Arundel County in the state of Maryland in the USA. Considering the analytic subgroups, the desired precision of estimates, and the available budget, it has been determined that these sample sizes are to be selected:
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 9. Designing Multistage Samples
Abstract
Previous chapters have covered the design of samples selected in a single stage. However, sampling is often done using more than one stage. There are a number of reasons why cluster or multistage sampling may be used. Using multistage samples can often be a practical and cost-efficient solution in situations where a list of elementary (or analytic) units is not available for direct sampling. In those cases, a list of elementary units can be compiled within just the sample clusters rather than for the whole frame. This is especially useful in household samples if a list of every household in a country, state, county, etc., is not available. In other cases, permission to do a survey may have to be obtained at the cluster level. For example, if the goal is to administer a standardized test to a sample of students, administrators in the school district or in the school may have to grant permission to do the survey.
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 10. Area Sampling
Abstract
Area sampling is a catchall term for a set of procedures in which geographic areas are selected as intermediate units on the way to sampling lower-level units that are the targets of a survey. Area sampling is just an example of multistage sampling, but because special data sources and methods are used, we devote a separate chapter to it. Calculations for determining sample allocations to the different stages are the same as those covered in Chap. 9.
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 11. The Area Sample Design: One Solution
Abstract
The project in Chap. 8 requested that you design a sample of twenty-five census tracts (m = 25) and one block group per sample census tract (n=1).
Richard Valliant, Jill A. Dever, Frauke Kreuter

Survey Weights and Analyses

Frontmatter
Chapter 12. Project 3: Weighting a Personnel Survey
Abstract
In this project you will develop survey weights and deliver an analysis file for a survey of military personnel. Members of the military reserves were asked a variety of questions about job satisfaction.
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 13. Basic Steps in Weighting
Abstract
Survey weights are a key component to producing population estimates.
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 14. Calibration and Other Uses of Auxiliary Data in Weighting
Abstract
The previous chapter described the first few steps used in weight calculation: base weights, adjustments of unknown eligibility, and nonresponse adjustments. The last step, which is extremely important in many surveys, is to use auxiliary data to correct coverage problems and to reduce standard errors. By auxiliary data, we mean information that is available for the entire frame or target population, either for each individual population unit or in aggregate form. These may be obtainable because a frame of all units in the population was used to select the sample and each listing on the frame contains some data. Surveys of business establishments or institutions may have such frames. Population totals for some variables may be available from a source separate from the survey, like a census. In a business survey, the frame might have the number of employees from an earlier time period for each establishment. In a household survey, counts of persons in groups defined by age, race/ethnicity, and gender may be published from a census or from population projections that are treated as highly accurate.
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 15. Variance Estimation
Abstract
In previous chapters we considered the variance of estimators in order to determine the sample size and allocation to the design strata. After the sample data are collected, estimates are made and their variances and standard errors (SEs) must be computed. An SE (square root of the estimated variance) is a basic measure of precision that can be used as a descriptive statistic, e.g., as part of a coefficient of variation (CV ), or for making inferences about population parameters via confidence intervals. Estimating SEs that faithfully reflect all sources of (or a significant portion of the) variability in a sample design and an estimator is our goal, but this can be complicated. This is especially true when several (random) weight adjustments described in Chaps 13 and 14 are used. For example, when an adjustment for nonresponse is applied and then weights are raked to population controls, both procedures contribute to the variance of an estimator in addition to the randomness due to selecting the initial sample itself.
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 16. Weighting the Personnel Survey: One Solution
Abstract
The project assigned in Chap. 12 was to compute a set of weights for a survey of members of the military reserves. A stratified simple random sample of personnel was selected and queried about satisfaction with their jobs. The project provides an opportunity to put into practice the techniques covered in Chaps. 13–15. Completing the project requires calculation of base weights, an adjustment to account for cases whose eligibility status is unknown, an adjustment for nonresponse, and calibration to some finite population totals. There are several practical problems to be solved, including selecting a particular method of nonresponse adjustment, deciding how to use the population counts that are available, and determining how to handle missing values in both the sample cases and the population counts.
Richard Valliant, Jill A. Dever, Frauke Kreuter

Other Topics

Frontmatter
Chapter 17. Multiphase Designs
Abstract
Sample designs are developed and estimators are chosen to efficiently fulfill specified analysis plans. Efficiency is generally defined to encompass three primary areas—accurate estimates (bias) with high levels of precision (small standard errors) calculated from data collected with procedures that make economical use of the study funds without exceeding the specified budget (cost). Sections 3.1 and 3.2 and Chap. 15 detail the gains achieved in precision if auxiliary information that is highly associated with the analysis variables can be used. This includes, for example, auxiliary variables used (i) in sampling as a stratification variable or to construct the measure of size for a probability proportional to size (pps) design or (ii) in estimation with a regression (or ratio) estimator. However, what if the only available sampling frame does not have useful auxiliary information? Without the auxiliary information, how might the statistician address concerns that the inflated sample size required for the specified level of precision will exceed the study budget?
Richard Valliant, Jill A. Dever, Frauke Kreuter
Chapter 18. Process Control and Quality Measures
Abstract
So far we have described a wide variety of tools and tasks necessary for sampling and weighting. Key to a successful project, however, is not only the mastery of the tools, and knowing which tool to use when, but also the monitoring of the actual process, as well as the careful documentation of the steps taken, and the possibility to replicate each of those steps. For any project, certain quality control measures should be taken prior to data collection during sample frame construction and sample selection and after data collection during editing, weight calculation, and database construction. Well-planned projects are designed so that quality control is possible during the data collection process and that steps to improve quality can be taken before the end of the data collection period. Obviously the specific quality control measures will vary by the type of project conducted. For example, repeated longitudinal data collection efforts allow comparisons to prior years, whereas one-time cross-sectional surveys often suffer from uncertainty with respect to procedures and outcomes. However, we have found a core set of tools to be useful for almost all survey designs and will introduce those in this chapter. We do want to emphasize that while it is tempting to think that assurance of reproducibility and good documentation is only worth the effort for complex surveys that will be repeated, in our experience, even the smallest survey “runs” better when the tools introduced here are used.
Richard Valliant, Jill A. Dever, Frauke Kreuter
Backmatter
Metadaten
Titel
Practical Tools for Designing and Weighting Survey Samples
verfasst von
Richard Valliant
Jill A. Dever
Frauke Kreuter
Copyright-Jahr
2013
Verlag
Springer New York
Electronic ISBN
978-1-4614-6449-5
Print ISBN
978-1-4614-6448-8
DOI
https://doi.org/10.1007/978-1-4614-6449-5