Statistical Properties of Construction Duration Data

TLDR

Construction simulators often approximate random process distributions with standard forms, yet the true underlying distribution is usually unknown and must be inferred from limited data or analogous processes. This study investigates the statistical properties of construction duration data to guide appropriate distribution selection. The authors analyzed 71 activity duration samples by plotting coefficient of skewness and kurtosis on a β1–β2 plane and by constructing histograms to compare each sample’s shape with known probability density functions. The analysis shows that flexible distributions such as the beta distribution and the Pearson system are necessary to capture the diverse characteristics of construction duration data.

Abstract

Simulators of construction operations often must approximate the underlying distribution of a random process using a standard statistical distribution (e.g., lognormal, normal, and beta). In many of these cases, the underlying distribution of the random process remains unknown to the modeler and its properties have to be inferred from the available sample of data or from experience with similar processes. This paper summarizes the findings of research aimed at investigating the statistical properties of construction duration data. Seventy‐one samples of durations of construction activities are analyzed by: (1) Plotting the sample's coefficient of skewness and kurtosis on a β1‐β2 plane; and (2) constructing histograms for each sample and comparing them with the shapes of known probability density functions. The results of the research indicate that flexible distributions (e.g., the beta distribution and Pearson system) are required to ensure proper modeling for the diversified characteristics of construction duration data.