Identifying component variability of end product specification tests

In recent years there has been a move away from methods and material specification for pavement projects and toward end product specifications (EPS) as part of a general improvement in quality assurance programs. Typically, the EPS are based on the average of a number of measurements on a given day, and the between-test variance and the between-days variance are ignored. This paper presents a methodology for identifying the components of the total variability, estimating the variance associated with each component, and identifying the responsibility of each component to overall variance. The total variability and individual component variability may be used for a number of quality assurance applications - for example, identifying potential problem areas in the construction of a particular pavement, as input to new end product specifications that account for the variability in the measured results, or for aiding in setting appropriate bonus/penalty schedules, which are typically used in conjunction with EPS. The statistical approach adopted in this paper is a nested analysis of variance and the methodology will be tested on five years of data from an EPS program in the Province of Alberta, Canada. Data are from Marshal density and field density tests of asphalt concrete, although the procedure may be used for any EPS field test.