# Confidence - Confidence Interval Problem: We want to...

This preview shows pages 1–9. Sign up to view the full content.

Confidence Interval Problem: We want to estimate the true mean of a random variable X accurately and economically. Approach 0: Estimating the mean with known variance Assumption 1) ( ) 2 ~, XN µσ 2) We know the variance 2 σ , but we do not know the mean µ . We want to estimate . Goodness of our assumption. Not practical, but a nice stepping stone toward a practical approach.

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

View Full Document
Step 1) Sample Mean Have a sample mean of size n , 1 1 ˆ n j j XX n = = where j X ’s are independent identical samples of X . Then ˆ [] EX µ = and 2 ˆ () VAR X n σ = . Step 2) Normalize Gaussian Sample Mean 2 ˆ , XN n    . Normalize the sample mean: ˆ ~ (0,1) / X N n
Step 3) Find the tail probability For an arbitrary 0 α > , find 2 z such that 2 2 2 2 1 2 2 u z Q z e du π   = =  . Then 2 ˆ P1 / X z n µ σ   ≤=   . 2 z

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

View Full Document
Step 4) the confidence interval Get 1 α confidence interval as follows. 22 2 2 2 ˆˆ ˆˆ // XX z z z Xz n n nn αα µ σσ −− ⇒− ≤ ≤ + implies ˆˆ Pr 1 µα   ≤≤ + =   ˆˆ , is called the 1 confidence interval   −+  . For a larger n , we will have a tight bound. 2 However we have to know the variance σ
Example. Assume ~ ( ,4) XN µ . We want to estimate . We will get the sample mean from 16 samples of X . Find the 80% confidence interval. For 80% confidence, 0.1 2 1 0.8 , 1.28 zz α −= = = . 2 Therefore 1.28 2/ 16 0.64 z n σ = ⋅= . We get ˆˆ Pr{ 0.64 0.64} 0.8 XX ≤≤ + = What does this mean? ˆ Go ahead to get 16 samples and the sample mean . ˆˆ It is true that 0.64 0.64 with probability 0.8 x xx ≤+

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

View Full Document
Using Excel 2 1 returns 2 NORMSINV z α    ref. Excel file ‘normal-dist Confidence Interval Example’ Error Function, Complementary Error Function Many books and computer libraries have tables or routines for finding error function 2 2 () () 1 z t o erf z e dt erfc z erf z π ≡− erf(z) is an integral of N (0,1/2) from -z to z . erfc : complementary error function. 1 ( ) (0,1) ( ) 2 2 z z Q z N dt erfc = = So 2 z is found from solving 2 2 z erfc = .
Review In many cases, it is safe to assume X is gaussian. However we can not expect to know the variance when we don’t know the mean

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

View Full Document
Sample Variance Assume 2 ~ (, ) XN µσ We want to estimate µ and we do not know 2 σ . Define the sample variance as 22 1 1 ˆ () 1 n j j s XX n = = , where j X ’s are independent samples of X , and ˆ X is the sample mean defined as before.
This is the end of the preview. Sign up to access the rest of the document.

## This note was uploaded on 11/23/2010 for the course EE EE528 taught by Professor Majungsoo during the Spring '10 term at Korea Advanced Institute of Science and Technology.

### Page1 / 27

Confidence - Confidence Interval Problem: We want to...

This preview shows document pages 1 - 9. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online