Likelihood principle¶
The likelihood priniciple states, that the inference should be based on the likelihood \(P(D|\theta)\) of the observed data, not based on hypothetical future dta that we have not observed.
The likelihood principle can be decomposed into two simpler principles:
Sufficiency principle This states that sufficient statistics contain all the relevant information about an unknown parameter (this is true by definition)
Weak conditionality This states that inference should be based on events that have happend and not on which might have happend.