Existing predictive equations for absolute durations (both bracketed and uniform) do not take into account the finite probability that durations of zero may be observed for a given set of predictor variables. Consequently, these equations overpredict the expected value of the duration. This note presents a methodology via which these zero values may be appropriately taken into account by treating duration as a censored semicontinuous random variable. The probability of observing a duration of zero is related to the value of peak ground acceleration for each record and this correlation is directly accounted for in the proposed methodology. The extent of overprediction is demonstrated through an example regression analysis on a subset of the Next Generation Attenuation (NGA) project dataset.