Model assisted survey estimation


We discuss two popular estimation methods that are often used in cluster and multistage sampling designs: Ratio and Regression estimation. The goal of these estimation methods is to make the estimation of a statistic of interest (e.g. mean) more efficient by using auxiliary variables. Inference is here not only using the inclusion probabilities anymore, but inference is being ‘assisted’ by the use of a statistical model. We introduce the idea of model-assisted inference.

Before class, make sure to finish last week’s take home exercise.

Additionally, you are required to prepare for the class discussion. For this week, find out what auxiliary data (at the sample level) is available for the survey you reviewed in week 40, and bring a list of auxiliary data to class. Think about about the following question (after reading the literature): would it be a good idea for your survey to use ratio/regression estimation if you are interested in estimating one of the variables of interest in your survey?


  • Lohr chapters 4, 7
  • Stuart (1984) 71-90





Take home exercise (for week 44)

Catch up on readings and exercises. Review parts of lectures/exercises again that you found difficult.