| Abstract: | Samples used to estimate discrete choice models in geography and regional science are typically assumed to be simple random samples. This assumption is not always met with existing samples. Furthermore, data collection is usually less costly if a stratified sampling strategy is adopted. Existing sampling theory suggests that stratified sampling can affect the consistency of parameter estimates. This paper reviews the salient points of this theory and relates theory to practice by means of a simple example on housing choice by elderly. |
