The document discusses the history and evolution of quantile regression as an approach to statistical analysis. It notes that traditional least squares regression focuses only on conditional means, but quantile regression allows exploration of other parts of the conditional distribution like quantiles. This offers a more comprehensive view beyond just the average. The document also states that while quantile regression is superfluous under the textbook linear model, in real applications it can reveal a less regular picture and heterogeneity in relationships. Finally, it expresses hope that quantile regression can help formulate more realistic models to ask more searching questions of data.
The document discusses the history and evolution of quantile regression as an approach to statistical analysis. It notes that traditional least squares regression focuses only on conditional means, but quantile regression allows exploration of other parts of the conditional distribution like quantiles. This offers a more comprehensive view beyond just the average. The document also states that while quantile regression is superfluous under the textbook linear model, in real applications it can reveal a less regular picture and heterogeneity in relationships. Finally, it expresses hope that quantile regression can help formulate more realistic models to ask more searching questions of data.
The document discusses the history and evolution of quantile regression as an approach to statistical analysis. It notes that traditional least squares regression focuses only on conditional means, but quantile regression allows exploration of other parts of the conditional distribution like quantiles. This offers a more comprehensive view beyond just the average. The document also states that while quantile regression is superfluous under the textbook linear model, in real applications it can reveal a less regular picture and heterogeneity in relationships. Finally, it expresses hope that quantile regression can help formulate more realistic models to ask more searching questions of data.
Much of the early history of social statistics, strongly influenced by Quetelet,
can be viewed as a search for the “average man” – that improbable man without qualities who could be comfortable with his feet in the ice chest and his hands in the oven. Some of this obsession can be attributed to the seductive appeal of the Gaussian law of errors. Everyone, as Poincaré famously quipped, believes in the normal law of errors: the theorists because they believe it is an empirical fact, and the empiricists because they believe that it is a mathematical theorem. Once in the grip of this Gaussian faith, it suffices to learn about means. But sufficiency, despite all its mathematical elegance, should be tempered by a skeptical empiricism: a willingness to peer occasionally outside the cathedral of mathematics and see the world in all its diversity. There have been many prominent statistical voices who, like Galton, reveled in the heterogeneity of statistical life – who resisted proposals to throw the mountains of Switzerland into its lakes. Edgeworth (1920) mocked excessive reliance on “reasoning with the aid of the gens d’arme’s hat – from which, as from the conjuror’s, so much can be extracted.” Models for the conditional mean in which independently and identically distributed Gaussian “errors” are tacked on almost as an afterthought are rife throughout the realms of science. They are indispensable approximations in many settings. We have argued that it is sometimes useful to deconstruct these models, complementing the estimation of models for the conditional mean with estimates of a family of conditional quantile functions. Under the idealized conditions of the textbook linear regression model, this step is quite superfluous. The conditional quantile functions all line up nicely parallel to one another and spaced according to the well-tabulated curve of the gens d’armes hat. If, however – and we have seen that this is not unusual – the quantile functions portray a less-regimented picture, then a deeper view into the data is revealed. Conditioning covariates may well shift the location, the central tendency, of the distribution of the response variable, but they may also alter its scale or change its entire shape. Quantile regression is intended to explore these possibilities. Gradually it is evolving into comprehensive approach to estimation and inference for models of
https://doi.org/10.1017/CBO9780511754098.010 Published online by Cambridge University Press
294 Quantile Regression
conditional quantile functions. Linear and nonlinear parametric models as well
as a variety of nonparametric models have been explored. By supplementing the exclusive focus of least-squares-based methods on estimation and inference about conditional means, quantile regression offers a view beyond the average man – a way to explore the sources of heterogeneity in statistical relationships. Having struggled with these ideas for nearly 30 years, it is very gratifying to find that they are gradually making their way into empirical applications across a wide range of scientific disciplines. There are still many unsettled questions and much unexplored territory. Completing the regression picture is an ambitious undertaking, one that can only be brought to fruition with the help of many hands. In the summer of 1959, L. J. Savage spoke to the Joint Statistics Seminar of Birkbeck and Imperial Colleges on the application of subjective probability in statistics. Savage’s talk and the ensuing discussion constitute a fascinating snapshot of statistical thinking of that time. D. R. Cox (1962) concluded his contribution to the formal discussion by saying: A final general comment is that the discussion above is of the question of how to reach conclusions about parameters in a model on which we are agreed. It seems to me, however, that a more important matter is how to formulate more realistic models that will enable scientifically more searching questions to be asked of data. (p. 53)
I would like to think that quantile regression could contribute to this important objective.
https://doi.org/10.1017/CBO9780511754098.010 Published online by Cambridge University Press