Comment: Struggles with Survey Weighting and Regression Modeling

Abstract

We congratulate the author on an informative andthought-provoking discussion on a topic of broadinterest to the statistics community: the fitting ofmodels to data collected through complex surveys.The number of papers written on this topic, whetherfrom a model-based or design-based perspective, issubstantial and goes back at least to Konijn (1962).This topic has led to some disagreements betweenthose advocating that the design best be ignoredwhen the primary interest is on the characteristics ofthe model, and those stating that the design cannotbe ignored. More recently, both sides of this discus-sion have moved to something approaching a con-sensus, with those favoring a model-based approachacknowledging the need to account for nonignorabledesigns in the model fitting, while the traditionaldesign-based view has been extended to explore cer-tain circumstances under which it is appropriate toignore the design.Thecurrentarticle isan excellent example of thoserecent discussions of why the design needs to be ac-counted for in modeling, and how this can be donein practice. The importance of