slides_7_converge

0 we started with the premise n a n o p 1 59 we have

This preview shows pages 59–65. Sign up to view the full content.

0. We started with the premise n a n o p 1 . 59

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

View Full Document
We have similar definitions for being bounded in probability. If n W n is bounded in probability we write W n O p n which is read “ W n is of order no greater than n in probability.” The definition applies to nonrandom sequences. The notation O p 1 again comes with 0. The definitions imply that if W n o p n then W n O p n . In practice we need relatively little of the o p / O p apparatus. It is useful for theoretical work, but we will mainly use o p 1 , O p 1 , O p n 1/2 , and O p n 1/2 . 60
EXAMPLE : Earlier we considered a standardized partial sum. We can appply the O p apparatus to the partial sum and sample average. Let X i : i 1,2,. .. be random variables with finite second moment that are pairwise uncorrelated, and assume E X i 0 and Var X i E X i 2 2 .Let Y n be the partial sum and X ̄ n Y n / n be the sample average. Using Chebyshev’s inequality we showed that n 1/2 Y n O p 1 which means Y n O p n 1/2 . We can also write this as n 1/2 X ̄ n O p 1 ,which means X ̄ n O p n 1/2 . 61

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

View Full Document
Note that when Y n is the partial sum, we can also write Y n o p n 1/2 for any 0, that is, n 1/2 Y n p 0. On the other hand, Y n O p n 1/2 for and 0. In other words, n 1/2 is the function of n that we must divide by in order to obtain a sequence that is bounded but does not have a variance shrinking to zero. We will make a stronger statement about such sequences in the next section. If E X i 0, we need to use n 1/2 X ̄ n n 1/2 Y n n as the sequence that is O p 1 . 62
6 . Convergence in Distribution Convergence in distribution is important for approximating probabilities involving averages and functions of them. It is very important for approximating the sampling distributions of estimators in a variety of settings (later). Consider the following general problem. We assume X i : i 1,2,. .. is a sequence of i.i.d. random variables with finite second moment. Let E X i , 2 Var X i .Let X ̄ n be the sample average of the first n variables in the sequence. 63

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

View Full Document
We immediately know the first and second moments of X ̄ n : E X ̄ n Var X ̄ n 2 / n If the common population distribution is Normal , 2 , then we know X ̄ n ~ Normal , 2 / n . But what if we do not know the population distribution (distribution of each X i ), or it is too complicated to be able to find D X ̄ n ? It is very helpful to be able to approximate the distribution of X ̄ n for a broad range of population distributions.
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

Page59 / 88

0 We started with the premise n a n o p 1 59 We have...

This preview shows document pages 59 - 65. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online