Skip to content

beta distribution isn't sufficiently justified #16

Description

@jack-mango

According to the maximum entropy principle a uniform distribution should be used if all you know is your parameter falls in some range and its a location type parameter. If the mean is constrained (which it seems it is based on your prior for # Gives contrast roughly 20% C = pm.Beta("C", alpha=20, beta=80)), you should use a truncated exponential distribution - as that's the maximum entropy distribution given your constraints.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions