ABSTRACT
In the statistical and scientific literature as a whole, there is a huge predilection for working with identified models, given their intuitive and appealing properties. As more data are collected, we become more and more certain about the values of target quantities. Moreover, we have clear mathematical understanding about this march to certainty: the extent of our uncertainty scales inversely with the square root of the sample size. No wonder then, investigators might not always want to fully ponder whether the assumptions behind an identified model are appropriate for the subject-area problem at hand. In fact, it is common to encounter scientific articles in which all quantitative analysis is carried out using an identified model, with qualitative concern about the model assumptions raised only in passing, perhaps in the discussion section of the paper. Arguably this is too late, since the results from the identified model analysis have already been imbued as the main findings of the work.
