Calibrating Visual Testing Devices Against One Another: The Pitfalls of Regression Analysis

摘要

When comparing data from instruments that measure related visual dimensions or in which there is inherent measurement error, simple linear regression creates a misleading representation of the relationship. Standard major axis analysis provides a best-fit line for comparing the two similar measures that is appropriate for two measured (dependent) variables, has intransitivity when axes are changed, is unaffected by changes in units or linear transformations, and is more reliable from one measurement session to another. This analysis allows manufacturers to more accurately relate their instruments to others and allows scientists to create models based on the best-fit lines relating two measured variables.