Project CEMAPRE internal
|Title||Identification of (nonlinear) models under Response Based Sampling|
|Participants||Pierre Hoonhout (Principal Investigator)|
|Summary||There is a wide range of questions that can only be answered using nonlinear regression models.|
Identification of nonlinear regression models is fairly well established under random sampling. For
some estimation problems, random sampling is not feasible. For instance, when investigating rare
diseases or the homeless a random sample is unlikely to contain sampling units that have the rare
disease or are homeless. In these cases, a response based sampling scheme is more appropriate.
Regression models are no longer identified without some kind of auxiliary information. The
identification results that are currently available in the literature are very model specific (e.g.
Imbens and Lancaster 1992, Econometrica). We derive identification conditions that cover any