Power and sample size calculations for comparison of two regression lines with heterogeneous variances.

ABSTRACT: The existence of interactive effects of a dichotomous treatment variable on the relationship between the continuous predictor and response variables is an essential issue in biological and medical sciences. Also, considerable attention has been devoted to raising awareness of the often-untenable assumption of homogeneous error variance among treatment groups. Although the procedures for detecting interactions between treatment and predictor variables are well documented in the literature, the corresponding problem of power and sample size calculations has received relatively little attention. In order to facilitate interaction design planning, this article describes power and sample size procedures for the extended Welch test of difference between two regression slopes under heterogeneity

SUBMITTER: Shieh G 

PROVIDER: S-EPMC6296670 | biostudies-literature | 2018

REPOSITORIES: biostudies-literature