Analysis of the Beta Distribution Assumption in the Pert Algorithm

In order to validate the assumption of beta distributed activity durations in the PERT algorithm, a study has been performed using three typical distributions: uniform, normal, and exponential. Random data were generated according to each of the three distributions to represent activity durations in three textbook example networks. The chosen networks had significant interaction between activities and numerous possible critical paths. PERT analyses were performed using two methods: (1) the true duration means and variances for each activity were used, and (2) the means and variances used were calculated assuming that the durations followed a beta distribution. The results of the two methods were compared to determine the similarity between the critical paths, the expected project durations, the project duration variances, and the durations required to give 95% probability that a project would be completed on schedule. The study was performed assuming that every activity in the network followed the same distribution. The results showed that there were variations in all areas analyzed. The average difference at the 95% confidence level was 8.75% for the uniform distribution, 3.62% for the normal distribution, and 16.12% for the exponential distribution