I've read the Wikpedia article Three-Points Estimation and the book "Software Estimation" written by Steve McConnell.
I globally grasped the process, however I have very little knowledge in statistics and I would like to be sure to understand some computations and approximations.
- PERT standard deviation
The following PERT formulae are used
expectedCase = (bestCase + 4 * mostLikelyCase + worstCase) / 6 standardDeviation = (worstCase — bestCase) / 6
If I understand correctly, the "6" number in the second formula assumes PERT follows a Normal Distribution and has nothing to do with the "6" in the first formula (which is a weighted mean). However PERT and software tasks rather follow a Beta Distribution.
Is there - and do estimators gerally use - a better formula following a Beta distribution?
In which cases must we be careful of this approximation, and NOT use the above standard deviation formula?
- Confidence interval
Once the expected value and standard deviation for the total project work time are computed, do we assume it follows a normal distribution due to the Central Limit Theorem?
It may explain why we need 20 tasks at least, and also why we can compute the confidence interval of the project using normal distribution formulae.