Models in survey sampling

Abstract

Models, especially in the form of assumed relationships between study variables and auxiliary variables, have influenced survey sampling theory and practice over the last four decades. Some of the early debates between the design-based school and the model-based school are revisited. In their pure forms, they offer two fundamentally different outlooks and approaches to inference in sample surveys. Complete reconciliation and agreement cannot be expected. But the tendency today is that each of the two approaches recognizes and profits from important elements in the other. We see an often fruitful interaction, as discussed in this article.