Comparison of Methods for Sensitivity and Uncertainty Analysis of Signalized Intersections Analyzed with the Highway Capacity Manual

Abstract

Most users of the Highway Capacity Manual (HCM) would prefer to have confidence intervals around the estimate of delay, but no procedure measures the uncertainty in delay and level of service. Four sensitivity analysis methods-partial differential analysis (PDA), partial correlation coefficient analysis, standardized regression coefficient analysis, and the Fourier amplitude sensitivity test (FAST)-and four uncertainty analysis methods-first-order analysis (FOA), Monte Carlo simulation (MCS), FAST, and the point estimate method (PEM)-were investigated. They were applied to data from an actuated signalized intersection. All input variables in the delay model except for the duration of analysis period were considered uncertain, for consistency with HCM. Day-to-day variation was the source of errors. Progression factor, cycle length, green time, and saturation flow are the most sensitive parameters. The incremental delay and upstream metering factors are the least sensitive. Volume and peak hour factor fall in between. The four uncertainty methods produced similar results for the mean, standard deviation, and confidence intervals of control delay for the base case. When the standard deviations of input parameters were doubled, MCS, FAST, and PEM produced similar results. PDA and FOA appear to be less suitable for sensitivity and uncertainty analysis, respectively, of the HCM delay model for signalized intersections.