Abstract

This paper examines the assumption of linearity in the application of regression techniques to tread wear data. Data from several sources provide evidence that tread wear is only approximately linear. Linear models appear to be adequate after a sufficient break‐in period whose length depends on the nature of the test. When linear models are used to compare the tread life potential of a test tire to that of a control tire, both tires must have had the same testing exposure or else the rate of wear may be overestimated.