Dorofeev S., Grant P. Statistics for Real-Life Sample Surveys: Non-Simple-Random Samples and Weighted Data

Cambridge: Cambridge University Press, 2006. — 266 p. — ISBN 978-0-521-85803-8.Samples used in social and commercial surveys, especially of the general population, are usually less random (often by design) than many people using them realise. Unless it is understood, this 'non-randomness' can compromise the conclusions drawn from the data. This book introduces the challenges posed by less-than-perfect samples, giving background knowledge and practical guidance for those who have to deal with them. It explains why samples are, and sometimes should be, non-random in the first place; how to assess the degree of non-randomness; when correction by weighting is appropriate and how to apply it; and how the statistical treatment of these samples must be adapted. Extended data examples show the techniques at work. This is a book for practising researchers. It is a reference for the methods and formulae needed to deal with commonly encountered situations and, above all, a source of realistic and implementable solutions.Sampling methods
Accuracy and precision
Design effect and sample size
Defining the objectives
Defining the constraints
Defining the population
Sampling frames
Simple random sampling
Multi-stage sampling
Stratification
Post-stratification
Systematic (interval) sampling
Cluster sampling
Complex samples
Selection with unequal probability
Non-response
Quota sampling
Other non-probability sampling methods
Sampling small populations
Some final thoughts
Weighting
Principles
Methods of weighting samples
Cell weighting
Marginal weighting
Aggregate weighting
Multi-stage and hybrid weighting systems
Statistical considerations
Ethical considerations
Statistical effects of sampling and weighting
Calculation of variance
Design effect and effective sample size
Weighting effect and calibrated sample size
Missing data
Significance testing
The purpose and philosophy of significance testing
Principles of significance testing
Choosing a test
Testing particular values of a mean or proportion
Testing the difference between means or proportions
The chi-square test of fit
Testing homogeneity and independence
Measuring relationships between variables
Analysis of contingency tables
Multivariate techniques
Regression analysis for weighted data
Classification and regression trees
Cluster analysis for categorical variables
Data fusion
Appendix A: Review of general terminology
Appendix B: Further reading
Appendix C: Summary tables for several common distributions
Appendix D: Chapter 2 mathematical proofs
Appendix E: Chapter 3 mathematical proofs
Appendix F: Chapter 4 mathematical proofs
Appendix G: Chapter 5 mathematical proofs
Appendix H: Statistical tables