You are on page 1of 45

Highlight the last lecture

The null hypothesis is the presumed condition that is accepted unless there is strong evidence against it. The alternative hypothesis is the claim that the researcher would like to establish based on the data, sometimes called a research hypothesis. The researcher would like to prove the claim under by rejecting However, the decision of not rejecting does not prove that is true

18 October 2011

STAT 101 -- Part IX

18 October 2011

STAT 101 -- Part IX

Equivalence of critical method and p-value method

p-value

Test statistics t

18 October 2011

STAT 101 -- Part IX

18 October 2011

STAT 101 -- Part IX

18 October 2011

STAT 101 -- Part IX

Number of Students made a mistake in each question


Question 14 Question 13 Question 12 Question 11 Question 10 Question 9 Question 8

Question 7
Question 6 Question 5 Question 4 Question 3 Question 2 Question 1 0 5 10 15 20 25 30 35 40 45 50

18 October 2011

STAT 101 -- Part IX

18 October 2011

STAT 101 -- Part IX

Relationship between hypothesis testing and confidence intervals

18 October 2011

STAT 101 -- Part IX

Relationship between hypothesis testing and confidence intervals (contd)

18 October 2011

STAT 101 -- Part IX

One sample test for a binomial distribution

18 October 2011

STAT 101 -- Part IX

10

http://bcs.whfreeman.com/bps3e/content/cat_010/applets/testsignificance.html
18 October 2011 STAT 101 -- Part IX 11

Under critical value method

Under p-value method

18 October 2011

STAT 101 -- Part IX

12

Example

A 1999 General Accounting Office (GAO) study found that a third of the 23.4 million retirees 65 or older supplemented Medicare with some form of employer coverage (Wall Street Journal, June 26, 2002) Suppose that in a current study, a random sample of 500 retirees 65 or older indicated that 185 supplemented Medicare with some form of employer coverage. At the 0.05 level of significance, is there evidence that the proportion of retirees 65 or older that supplement Medicare with some form of employer coverage is now different from one-third?

p-value=0.0409*2=0.0818

18 October 2011

STAT 101 -- Part IX

13

18 October 2011

STAT 101 -- Part IX

14

Discussion of example A 95% confidence interval for p is (0.3277, 0.4123) Since the confidence interval includes 0.3333, it implies that the null hypothesis of p=0.3333 is not rejected based on confidence interval approach Noted that the confidence approach is not always consistent with the hypothesis testing approach.

18 October 2011

STAT 101 -- Part IX

15

IX: Two-Sample Tests


Test

the difference between two independent population means (standard deviations known or unknown) Test two means from related samples for the mean difference Test the difference between two population proportions
18 October 2011 STAT 101 -- Part IX 16

Two sample tests


Two Sample Tests

Test Population Means, Independent Samples

Test Population Means, Related Samples

Test Two Population Proportions

18 October 2011

STAT 101 -- Part IX

17

Difference between two population means

Test hypotheses for the difference between two population means Different data sources: samples are unrelated and independent Use the difference between two sample means as the point estimate Use the Z test statistic if both population variances are known Use the pooled variance t-test if both population variances are unknown

18 October 2011

STAT 101 -- Part IX

18

Two samples test: both population variances are known

Assumptions: Samples are randomly and independently drawn Population distributions are normal or both sample size are greater than 30 (by using CLT) Population standard deviations are known The test statistic for

18 October 2011

STAT 101 -- Part IX

19

Hypothesis tests for two population means

Two population means, Independent samples Two-tail test

18 October 2011

STAT 101 -- Part IX

20

Two samples test: both population variances are unknown


Assumptions: Samples are randomly and independently drawn Population distributions are normal or both sample size are greater than 30 (by using CLT) Population standard deviations are unknown but assumed equal

18 October 2011

STAT 101 -- Part IX

21

Estimating the variance

The population variances are assumed equal, so use the two sample standard deviations and pool them to estimate population variance The pooled standard deviation is

18 October 2011

STAT 101 -- Part IX

22

Test statistic

18 October 2011

STAT 101 -- Part IX

23

Numerical example

The Computer Anxiety Rating Scale (CARS) measures an individuals level of computer anxiety on a scale from 20 (no anxiety) to 100 (highest level of anxiety). Researchers at Miami University administered CARS to 172 business students. One of the objectives of the study was to determine if there is a difference between the level of computer anxiety experienced by female students and male students. At the 0.05 level of significance, is there evidence of a difference in the mean computer anxiety experienced by females and males?

18 October 2011

STAT 101 -- Part IX

24

18 October 2011

STAT 101 -- Part IX

25

18 October 2011

STAT 101 -- Part IX

26

Numerical example

A problem with a telephone line that prevents a customer from receiving or making calls is disconcerting to both the consumer and the telephone company. The data file phone.xls represent samples of 20 problems reported to two different offices of a telephone company and the time to clear these problems (in minutes) from the customers lines. Set the significance level 0.05.

18 October 2011

STAT 101 -- Part IX

27

18 October 2011

STAT 101 -- Part IX

28

Discussion:

A professor wants to investigate the text-book price differences between the campus bookstore and the competing off-campus bookstore. Approach I: The professor randomly selects 30 textbooks in campus bookstore and randomly selects other 30 text-books in the off-campus store. Approach II: The professor randomly selects 30 text-books which are sold in both stores and compares their price differences between the two bookstores.
STAT 101 -- Part IX 29

18 October 2011

Two sample test: related samples

Test the means of 2 related populations Paired or matched samples Repeated measures (before/after treatment) Eliminate variation among subjects Use the difference between paired values

Assume both populations are normal or (By CLT when sample sizes are large.)

18 October 2011

STAT 101 -- Part IX

30

Mean difference

18 October 2011

STAT 101 -- Part IX

31

Hypothesis testing for mean difference


Paired samples

18 October 2011

STAT 101 -- Part IX

32

Numerical example

Can students save money by buying their textbooks at Amazon.com? To investigate this possibility, a random sample of 15 textbooks used during the spring 2001 semester at Miami University was selected. The prices for these textbooks at both a local bookstore and through Amazon.com were recorded. At the 0.01 level of significance, is there evidence of difference between the mean price of textbooks at the local bookstore and Amazon.com?

18 October 2011

STAT 101 -- Part IX

33

18 October 2011

STAT 101 -- Part IX

34

18 October 2011

STAT 101 -- Part IX

35

Hypothesis test to two population proportions

Test a hypothesis for the difference between two population populations


The test statistic is normally distributed if the following conditions are satisfied:

18 October 2011

STAT 101 -- Part IX

36

18 October 2011

STAT 101 -- Part IX

37

Hypothesis tests for two population proportions

18 October 2011

STAT 101 -- Part IX

38

Numerical example

A sample of 500 shoppers was selected in a large metropolitan area to determine various information concerning consumer behavior. Among the questions asked was, Do you enjoy shopping for clothing? Of 240 males, 136 answered yes. Of 260 females, 224 answered yes. Is there evidence of a significant difference between males and females in the proportion who enjoy shopping for clothing at the 0.01 level of significance?

18 October 2011

STAT 101 -- Part IX

39

Since the sample size requirements are satisfied, the test statistic is approximately normal due to CLT.

18 October 2011

STAT 101 -- Part IX

40

18 October 2011

STAT 101 -- Part IX

41

Useful and interesting websites


http://bcs.whfreeman.com/bps3e/content/c at_010/applets/testsignificance.html Simulation of test significance

http://www.ruf.rice.edu/~lane/stat_sim/com pare_dist/index.html

Compare two populations experiment Robustness of two sample t-test

http://www.ruf.rice.edu/~lane/stat_sim/robu stness/index.html

http://www.amstat.org/publications/jse/java/v9n1/andersoncook/BadExpDesignApplet.html http://www.amstat.org/publications/jse/java/v9n1/andersoncook/GoodExpDesignApplet.html
18 October 2011 STAT 101 -- Part IX 42

Recommended questions from the textbook


Question 9.56 10.10; 10.14 10.20 10.24 10.32; 10.34 Page 352 372 380 381 388

18 October 2011

STAT 101 -- Part IX

43

Some Examples: Assignment 5 Projects


Observational Study (Descriptive Statistics) Go to a local grocery store and collect these data for at least 75 breakfast cereals: cereal name; grams of sugar per serving; and the shelf location (bottom, middle, or top). Group the data by shelf location and use three boxplots to compare the sugar content by shelf location

Experimental Study (Hypothesis Testing) Conduct a taste test of either Coke versus Pepsi or Diet Coke versus Diet Pepsi. Survey at least 50 randomly selected students who identify themselves beforehand as cola drinkers with a definite preference for one of the brands you are testing. Give each subject a cup of each cola that has been coded in a way known only to you. Calculate the fraction of your sample whose choice in the taste test matches the brand identified beforehand as their favorite. (Do not tell your subjects that this is a test of their ability to identify their favorite brand; tell them it is a test of which tastes better)

18 October 2011

STAT 101 -- Part IX

44

Observational Study (Confidence Interval Estimation) Estimate the average number of hours that students at SMU sleep each day, including both nighttime sleep and daytime naps. Also estimate the percentage who have been up all night without sleeping at least once during the current semester.

Observational Study (Two-sample Hypothesis Testing) What percentage of the seniors at SMU expect to be married within five years of graduation? What percentage expect to have children within five years of graduation? How many biological children do the seniors at your college expect to have during their lives? Do males and females differ in their answer to these questions?

18 October 2011

STAT 101 -- Part IX

45

You might also like