Choosing a probability distribution
Choose a distribution whose range and shape match what is known about a changing input.
On this page
A duration that is fixed by design or set somewhere else.
You know a hard minimum and maximum but have no reason to favour values inside them.
Positive memoryless waits with many short observations and occasional long ones.
Measurements cluster symmetrically around a mean and large deviations are uncommon.
Positive durations formed by several multiplicative effects, producing a right tail.
You can estimate a minimum, most-likely value, and maximum but have little raw data.
Positive right-skewed durations made from accumulated waiting or work stages.
Positive durations where completion or failure likelihood changes with elapsed time.
Heavy-tailed processes where rare extreme durations are part of system risk.
You have observed durations and want to preserve their measured shape.
Only a finite set of durations can occur and each has an observed or chosen weight.
A finite evenly spaced set of values is possible and each should be equally likely.
There are exactly two possible durations tied to a binary outcome.