Variable selection and transformation in linear regression models

초록

We develop a method for comparing separate linear models, for a common response variable that may be expressed on different scales and may be described by distinct explanatory variables. A method of stochastic simulation is used to approximate the fitted maximum likelihood estimates and then the Cox statistic is computed to test separate linear models. The bootstrap iteration is also used to calibrate confidence intervals to correct the test level. (c) 2005 Elsevier B.V. All rights reserved.

키워드