The penalty for assuming that a monotone regression is linear

DOI10.1214/AOS/1176350279MaRDI QIDQ1819508zbMATH OpenOpenAlexFDO

Authors David Fairley, Dennis Pearl, Joseph Verducci

Publication date 1987

Published in The Annals of Statistics (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1214/aos/1176350279

zbMATH Keywords

expected squared-error lossextralinear variationfixed marginsintrinsic variationlinear regressionmonotone regressionpenalty

Mathematics Subject Classification ID

Statistical distribution theory (62E99) General nonlinear regression (62J02)

Review text

For jointly distributed random variables (X,Y) having marginal distributions F and G with finite second moments and F continuous, the proportion of Var(Y) explained by linear regression is \([Corr(X,Y)]^ 2\) while the proportion explained by E(Y\(| X)\) can be arbitrarily near 1. However, if the true regression, E(Y\(| X)\), is monotone, then the proportion of Var(Y) it explains is at most \(Corr[Y,G^{-1}(F(X))]\).

Recommendations

• On monotonicity of regression quantile functions
• A linear fit gets the correct monotonicity directions
• Monotonicity of regression functions in structural measurement error models
• Monotone nonparametric regression with random design
• scientific article; zbMATH DE number 1304679

This page was built for publication: The penalty for assuming that a monotone regression is linear

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1819508)