A Recursive Method of Transforming a Response Variable for Linearity

  • Seo, Han-Son (Dept. of Applied Statistics, Kon-Kuk University)
  • Published : 1998.08.01

Abstract

We consider a graphical method for visualizing the strictly monotonic transformation of t(y) so that the regression function E(t(y)$\mid$x) is linear in the predictor vector x. Cook and Weisberg (1994) proposed an inverse response plot which relies on the results of Li & Duan (1989) to obtain consistent estimates. Based on the recursive addition of the results from the two dimensional plots, we propose a new procedure which can be used when the consistency result is in doubt.

Keywords

References

  1. Journal of Royal Statistical Society, B v.35 Testing transformations to normality Atkinson, A. C.
  2. Journal of Royal Statistical Society, B v.44 Regression Diagnostics, transformations and Constructed Variables (with discussion) Atkinson, A. C.
  3. Plots, Transformations and Regression Atkinson, A. C.
  4. Journal of Royal Statistical Society, B v.26 An Analysis of Transformations (with discussion) Box, G. E.;Cox, D. R.
  5. Journal of Multiplicate Analysis v.7 On the Theory of Elliptically Contoured Distributions Cambanis, S.;Huang, S.;Simons, G.
  6. Regression Diagnostics With Dynamic Graphics Technometrics v.31 Cook, R. D.;Weisberg, S.
  7. Three Dimensional Residual Plots Proceedings of the 21st Symposium on the Interface: Computing Science and Statistics Cook, R. D.;Weisberg, S.
  8. Biometrika v.81 Transforming a response variable for linearity Cook, R. D.;Weisberg, S.
  9. Journal of Multivariate Analysis v.20 A Characterization of Spherical Distributions Eaton, M. L.
  10. Annals of Statistics v.17 Regression Analysis under Link Violation Li, K. C.;Duan, N.
  11. LISP-STAT Tierney, L.