This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: 22 Estimating the population proportion l Recall binomial rv X l Number of successes in n trials l Know that E( X )= np & Var( X )= np(1p) l Now consider the rv X / n l This is simply the sample proportion of successes l It is an unbiased estimator of the population proportion l Why? l It has a variance of p(1p)/n l Why? 23 Estimating the population proportion… l Also recall that for large n, X is approximately normal l Thus the sample proportion is also approximately normal for large n using the CLT l Approximate sampling distribution follows immediately l As do confidence intervals & hypothesis tests ) 1 , ( ~ / ) 1 ( ˆ and ) 1 ( , ˆ then ˆ Let N n p p p P Z n p p p ~N P n X P =  = Predicting election results l How do political commentators predict election results so soon after polls close & before all votes are counted? l They use data from exit polls and early returns as samples l Conduct inference on true (as yet unknown) proportion of voters supporting the incumbent A l In a 2 candidate race an exit poll of 400 produces 212 people who say they voted for A l What do you conclude about A’s reelection? 24 25 Predicting election results... l Set the problem up as an hypothesis test l A will win if s/he gains more than 50% of the vote & will lose if <50% of votes are cast in his/her favour l So p= 0.5 is the limit, above which s/he will win; this sets our null hypothesis l Is there sufficient evidence to refute this null? ! ! call to close too error, of margin within 1151 . ) 2 . 1 ( 2 . 1 400 / 25 . 5 . 53 . and 400 5 . 5 . , 5 . ˆ then .53 400 212 ˆ 5 . : 5 . : 1 ⇒ = = = = × = = = Z P value p Z ~N P P p H p H 26 Predicting election results… l In hypothesis testing example assumed H to be true l Allows calculation of the variance of the sampling distribution l What happens when constructing a confidence interval for p? l Use sample proportion to estimate p l Test statistic will remain approximately normal than.5 less are that of values includes CI 98% our .47,.59) ( or 400 / ) 47 . 53 (. 33 . 2 53 . is for CI 98% the & 33 . 2 , .02 For / ˆ 1 ˆ ˆ 01 . 2 p p z n ) P ( P z P α/ ⇒ × × ± = = ± α 27 Progress report #6 l Have the statistical tools that underpin a large range of statistical procedures l Know about some key distributions, eg Binomial, Uniform, Normal & t l Introduced basics of point & interval estimation l Introduced basics of constructing & conducting hypothesis tests l All of this has been done in a univariate context l Most interesting problems involve multivariate situations where relationships between variables is important l Takes us back to correlation & regression l Previously introduced as descriptive devices l Now consider inference in a regression context...
View
Full Document
 One '11
 DenzilGFiebig
 Normal Distribution

Click to edit the document details