Sampling Design & Weighting

Gathering information about the characteristics of a population for each of its members is time consuming and costly. However, we can make inferences about the population with the help of samples. Sampling is a statistical procedure that is concerned with the collection of data on a part of a population of interest and subsequent extrapolation to make statistical inferences about the whole population, for example using sample means to estimate population means. Sampling is essential to enable collection the vast amount of information required for understanding the functioning of our society. It provides a solid basis for estimating unknown values and ratios, and for testing the validity of presumed relationships in different areas of science. At the same time, it is crucial in assuring the efficient use of scarce resources for data collection, while maintaining data quality. One important aspect of probability sampling is its ability to produce not only valid estimates of the parameters of interest but also, under broad conditions, of their margin of errors due to sample variability. This feature has no doubt greatly contributed to the wide acceptance of probability sampling as an objective tool of measurement among analysts and the public at large. OBJECTIVES The main objective of the course is to “enhance understanding and capacities of ILO constituents and social partners to design and implement household surveys and to process sample data in line with best methodological practices.” The course will enhance the knowledge of participants on the different sampling and weighting techniques highlighting their pros and cons. In addition, the course will highlight the link between sampling techniques and survey design with a particular focus on labour force surveys (LFS), the most common source of official labour statistics across the globe. More specifically, the course aims to: • Enhance understanding on sample surveys and survey designs; • Provide insights about the principles and practices of sampling • Enrich understanding of estimation theory, methods for probability sampling, and sampling frames; • Improve understanding of different weighting strategies and treatment of unit non-response; • Increase understanding of quality dimensions and calculation of sample size for complex multi stage designs • Provide practical case studies on the treatment of total non-response and on weighting making use of different sets of benchmarks available for different population sub-groups and/or for different geographical domains. CONTENT • Overview of sample surveys • Overview of survey designs • Introduction to principles and practices of sampling • Estimation Theory • Sampling frames • Use of master samples for household surveysSampling designs and methods for probability sampling (advanced) • Selecting samples in complex designs with Probability Proportional to Size • Sampling rotation • Weighting strategies • Sampling strategy for LFS with migration modules • Software for calibration: R and the package ReGenesees • Post-stratification and calibration: Examples and case studies using ReGenesees • Calibration: Weighting with complex calibration constraints on several layers and sub-groups (households and individuals, national and non-nationals, urban and rural, different geographical domains of estimation, etc.) using ReGenesees • Unit Non-Response: Non-response assessment (using logistic models/regression trees) and treatment (using non-response adjustment cells) Treatment of Unit Non-Response (Examples and Case Studies) • Quality Dimensions • Calculation of sample size for complex multi stage designs • Use of an R package to calculate standard errors and confidence intervals (for levels, ratios and rates), and design effect • Use of an R package to calculate standard errors and confidence intervals (for levels, ratios and rates), and design effect • Use of an R package to calculate sample size given the desired precision, budget and other constraints (Examples and Case Studies) ACHIEVEMENT Criteria is scoring more than 60% in the final exam. NUMBER OF HOURS: 60