Analysing Complex Survey and Subsample Data (with R)

Modelling data from complex sampling designs such as multistage surveys or case-cohort designs used to require specialised expertise and software (and even hardware). In the modern world, commodity computers, free software, internet data distribution and the unifying ideas of weighting and influence functions mean that anyone can analyse datasets like these, and relatively little expertise is needed. This workshop is designed to provide a basis for doing your own analyses. We will cover

• the basic features of sampling: strata, clusters, sampling probabilities
• describing sampling to R
• summary statistics and graphics for survey data
• regression modelling for survey data
• calibration/raking of weights to use population or full-cohort information
• efficiency/robustness issues in weighted estimation

The workshop will alternate between presentations by me and exercises for you to work on.

Requirements

Assumes some experience using R, though we won’t need any advanced programming, and knowledge of regression modelling at an advanced undergraduate level.  You will need a laptop with R, the survey package, and the svyVGAM package installed.

Biography

Thomas Lumley is a Professor in the Statistics Department at the University of Auckland and Affiliate Professor at the University of Washington.  He is the developer of the ‘survey’ package for R and author of “Complex Surveys: A Guide to Analysis Using R” (Wiley, 2010). Thomas has given workshops on R or surveys in 13 different time zones.

Cancellations policy

a. Registration cancellations must be made in writing to the Chair of the Local Organising Committee (emi.tanaka@anu.edu.au) for consideration by the committee. 

b. Cancellations after November 8, 2025 may incur some penalty.