Performance of Ols-Bisector Regression in Method Comparison Studies

Abstract

Regression Analysis is a widely used technique in method comparison studies. There are several studies about estimating and testing interactions in a linear regression model but It's very important to use correct regression technique in comparison study to obtain correct results. In method comparison studies, when both of the dependent and the independent variables include some measurements errors, Type II Regression techniques must be used to calculate the correct parameters. The aim of this study is to discuss eight different regression techniques (OLS, OLS-Bisector, Major Axis (MA), Reduced Major Axis (RMA), Deming, Pas sing -Bab lok, York and Theil) that may be used in method comparison studies and which are the alternatives of Ordinary Least Squares (OLS) regression analysis when the assumptions of OLS are not met and to suggest alternative techniques for calculating the correct linear relationship between the two methods. In simulation part of this study there has been generated different types of data and in all conditions, OLS-Bisector method, which bisects the OLS (Y/X) and OLS (X/Y), estimated the parameters near to real values and, show the best performance then all other techniques. © IDOSI Publications, 2011.