64 but we must be more formal than just saying

Info iconThis preview shows pages 64–72. Sign up to view the full content.

View Full Document Right Arrow Icon
64
Background image of page 64

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

View Full Document Right Arrow Icon
But we must be more formal than just saying “approximate the distribution of X ̄ n .” If we simply study the distribution of X ̄ n as n we get nothing useful: X ̄ n p which means the distribution of X ̄ n collapses to the single point with probability one. (Called point mass at .) One way to see this is that Var X ̄ n 2 / n 0as n . The next graph shows 3. The limiting distribution, which of course is discrete, is a very poor approximation to the actual distribution of X ̄ n even when n is very large. 65
Background image of page 65
n = 120 n = 50 n = 25 limiting distribution 0 1 2 3 4 3 Distributions of Sample Average with Mean = 3 66
Background image of page 66

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

View Full Document Right Arrow Icon
The problem is that we are using the wrong sequence of random variables. We need to standardize X ̄ n so that the variance does not shrink to zero. If we use Z n X ̄ n E X ̄ n  SD X ̄ n X ̄ n / n n X ̄ n then E Z n 0 and Var Z n 1. The distribution of Z n can be virtually anything, but we know the distribution is not collapsing as n . In the next section we will show that the distribution of Z n settles down to a standard normal. 67
Background image of page 67
DEFINITION : The sequence of random variables W n : n 1,2,. .. converges in distribtion to the random variable W if F n w F w as n for all w where F is continuous. Here, F n w P W n w and F w P W w , where W has distribution F . We are actually talking about (pointwise) convergence of a sequence of CDFs to a given CDF, but it is convenient to refer to a random variable with distribution F . We call F the limiting distribution or asymptotic distribution of W n . 68
Background image of page 68

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

View Full Document Right Arrow Icon
As notation, we write W n d W ( W n converges in distribution to W )or W n a ~ W ( W n is asymptotically distributed as W ). When we know W has a specific distribution, such as standard normal (the leading case), we might write W n d Normal 0,1 W n a ~ Normal 69
Background image of page 69
In the cases we are interested in, F  is continuous at all w .In addition to the normal, the chi-square distribution is an important case. In many applications, the F n  are not always continuous. In fact, important cases include when F n  has some jumps for all n (but the jumps become less and less pronounced as n increases, and so F n  converges to a continuous function). 70
Background image of page 70

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

View Full Document Right Arrow Icon
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 .
Background image of page 71
Image of page 72
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

Page64 / 88

64 But we must be more formal than just saying approximate...

This preview shows document pages 64 - 72. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online