slides_7_converge

70 continuous convergence theorem once we have

This preview shows pages 70–77. Sign up to view the full content.

70

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Continuous Convergence Theorem Once we have established convergence in distribution, we can derive the limiting distribution of lots of functions of the original series. THEOREM : Suppose W n d W . If g  is a continuous function on then g W n d g W EXAMPLE : Suppose we know Z n d Z ~ Normal 0,1 . Then Z n 2 d Z 2 ~ 1 2 and exp Z n d exp Z ~ Lognormal 0,1 . 71
Facts About Convergence in Probability , Boundedness in Probability , and Convergence in Distribution (i) If W n p W then W n d W . In practice, this is not especially useful because typically W c for some constant, and then W n d c , which means the limiting distribution is P W c 1. (ii) If W n d W then W n O p 1 . This is very useful because it implies that if we can show a sequence converges in distribution then it is automatically bounded in probability. We conclude immediately that if W n d W and g  is continuous then g W n O p 1 [because g W n d g W .] 72

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
(iii) The asymptotic equivalence lemma allows us to obtain the limiting distribution of one sequence if we know (a) that it is “getting close” to another sequence and (b) we know the limiting distribution of that other sequence. Formally, suppose Z n d Z and W n Z n p 0 or W n Z n o p 1 . Then W n d Z The result is used routinely for large-sample approximations to estimators and test statistics. 73
Convergence in Distribution for Random Vectors We can use the natural extension for random vectors. Namely, W n d W if F n w F w at all w k where F  is continuous. But there is also a useful equivalent condition based on convergence of linear combinations. FACT : W n d W if and only if for all a k with a a 1, a W n a 1 W n 1 a 2 W n 2 ... a k W nk d a W . 74

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
In other words, we can establish convergence of the joint CDF for a sequence of random vectors by establishing convergence of univariate linear combinations of the random vector. This characterization is especially useful for establishing convergence to multivariate normality because linear combinations of a multivariate normal are normal. The continuous convergence result continuous to hold. If W n d W and g : k m is continuous then g W n d g W . 75