Stratified and cluster sampling

Introduction

In practice, real surveys seldomly use a simple random sampling design. We will discuss why stratification is often used to make the sample more efficient (allowing smaller samples), and why clustering is used to limit survey costs.

Literature

• Stuart (1984) p. 35-71
• Neyman (1934)

Optional:

• Lohr, S. (2022), Sampling: design and analysis, chapter 3 (stratified sampling) and 5 (cluster sampling)

Lecture

We start by discussing the Take home exercise. After this, we illustrate the benefits of stratification for survey efficiency and clustering for survey costs using example data.
Slides

simulation stratified design

simulation cluster design

Class exercise

Specify a stratified and clustering design in R

data
R exercises
R exercise solutions

R exercise complete answers

Take home exercise

Review the survey documentation of your adopted design-based survey (see weeks 1,3). Work out the sampling design on paper (not in R!), and bring the results to class.
R exercises