It is common practice to apply a priori belief about the importance of variables on pavement wear as a measure of the model's success as an inferential tool. For example, it is commonly held that traffic causes pavement wear and hence a statistical result showing traffic as insignificant on pavement wear is considered a failure of the data, the modelling approach or both. This technical discussion attempts to raise the issue of multicollinearity and endogeneity and why variables such as traffic, pavement depth, climate etc. should not be considered a failure of the data or modelling approach if found insignificant. The authors wish to alter the current opinion of many in the field who define condition/time models as overly simplistic and unrealistic. The authors attempt to highlight that this may in fact be the result of a high quality pavement design procedure which needs acknowledging. A likely scenario of increasing quality in pavement design methods will in turn make identifying significant variables other than time from in-service pavement data more difficult. Further, the authors attempt to persuade readers with the opinion that including insignificant variables, simply due to the a priori engineering belief of their importance in a pavement deterioration model, is actually unhelpful to both the model and pavement management. (a).
Abstract