Sample Based Trip Length Distribution Quality Based on Chi-Square and Mean Absolute Error Value

Abstract

As a vital part of transport modelling, the trip length distribution is normally gotten from a sample. In order to get a good model, a method to test the quality of the sample based trip length distribution must be avaliable. A method to measure this quality has ever been developed, but it was found that the existing method still can be improved. While still using goodness of fit statistical test, the new method propose two quality measurements. First, to verify whether the observed trip length distribution is conforrm to the reference trip length distribution at a certain confidence level value, indicated by a value of . Second, to verify whether the error, mean absolute error measured in  percentage, |e%| = 100 × |e|_mean / x_i,mean is acceptable. The new method can be used, while still fulfilling the basic principle of sample quality measure, i.e. satisfying maximum acceptable error at a certain confidence level.