Author : Monika Buljan
Publisher :
ISBN 13 :
Total Pages : 66 pages
Book Rating : 4.:/5 (657 download)
Book Synopsis The Coverage Probablity of Confidence Intervals in Regression After a Test for Parallelism by : Monika Buljan
Download or read book The Coverage Probablity of Confidence Intervals in Regression After a Test for Parallelism written by Monika Buljan and published by . This book was released on 2009 with total page 66 pages. Available in PDF, EPUB and Kindle. Book excerpt: We consider an analysis of covariance with a single covariate and a number of treatments. It is common practice to perform a preliminary test of the null hypothesis of parallelism for each treatment group before proceeding with further inference. If the null hypothesis is accepted, that is, that the slope of each of the straight lines is the same, then further inference proceeds on the assumption that the slopes of the lines are actually the same. Of course it is possible that the particular slopes of the lines are different, even when the null hypothesis of parallelism is accepted. This thesis will focus on inference in the form of confidence intervals for treatment differences. We will specifically consider how a preliminary hypothesis test for parallelism affects the coverage probability of these confidence intervals. This is explored by conducting Monte-Carlo simulations based on a real life data set using the programming language R. The real life data are about the effect of additives on weight gain in animals, which has four treatments with three subjects in each treatment group. It turns out that performing an initial test of parallelism is problematic because the minimum coverage probability of the confidence intervals falls below the nominal coverage. Hence the confidence intervals constructed after a preliminary test of parallelism are in fact, misleading.