Chief Expert in Statistical Applications, PWD, Tel Aviv,
(Received 20 August 2003; accepted 17 March 2004)
Statistical behavior and peculiarities of the viscosity data for original and aged bituminous materials have been investigated. Approximation of the empirical distribution of the aging index values as the ratio of corresponding values for aged and original bituminous material viscosity has been carried out employing normal, lognormal, and Charlier theoretical distribution models. Corresponding models for the suitable approximation of values under consideration have been verified. Probability distribution models under discussion reflect the main features of the aging index data. Accordance of the applied distribution models has been confirmed by transformed-values analysis employing logarithmic and square root transformations for presented aging index values. Suggested probabilistic models including the normal distribution and, in the first instance, the lognormal and Charlier distribution models, can be recommended for practical application for corresponding data analysis. Relationships among the viscosity data of original and aged bituminous materials employing different linear and nonlinear equations and multiple regression models, also including use of penetration and kinematic viscosity data, have been studied and analyzed. Specifications of the residual behavior for the regression equations have been examined, keeping in mind data peculiarities and their degree of heterogeneity. Presented results widen the possibilities for original and aged bituminous materials penetration and aging index data interpretation, increasing the reliability of estimations for corresponding quantitative characteristics.
Paper ID: JTE12194