Consistency

Summary

Given:

  • A sample x1,x2,...,xnx1,x2,...,xn
  • A statistical model where θθ is an unknown parameter
  • An estimator ˆθn=g(x1,,xn)^θn=g(x1,,xn)

Definition: consistency

  • We may view ˆθ1,ˆθ2,ˆθ3,^θ1,^θ2,^θ3, as a sequence of random variables .
  • We say that ˆθn^θn is a consistent estimator of θθ if ˆθn^θn converges in probability to θθ ,

plim ˆθn=θplim ^θn=θ

  • or

P(|ˆθnθ|<ε)1 as n

  • for all ε>0 .

Example

  • x1,x2,...,xn is an IID random sample where E(xi)=μ and Var(xi)=σ2 . Then ˉxn is a consistent estimator of μ ( the weak law of large numbers ).