Transcript Document

CHAPTER 14

Chi- square goodness of fit Test “GOF”

Chi-square Test for Homogeneity

Chi-square Test for Independence DONE SAME WAY

Symbol for Chi-Square is

χ

2

What’s the difference?

Goodness of Fit…..how well does the observed data match the expected….2 rows or 2 columns • • Homogeneity…..More than one sample is taken with one categorical variable in mind

(2+ Samples, 1 category)

• • Independence/Association…..Only one sample is taken and there are two or more categories.

(1 sample, 2+ categories)

Chi-Square Curve

It is not a NORMAL CURVE!!!

It is always skewed to the right some

Is your die fair—one more time.

Roll your die 60 times.

Write down the number for every roll.

Chi-Square GOF Test

• If your die is fair you would expect to get 10 of each number in 60 rolls.

• In this test we compare the EXPECTED results vs the OBSERVED results.

Hypotheses

• Ho: The proportion of each number that occurs on my die is 1/6 • Ha: The proportion of each number that occurs on my die is different than 1/6 • There are no symbols for the chi-square. However, it is always one-sided, even though the word “ different ” is used.

Normal condition for

χ

2

• 80% of the expected cells are greater than or equal to 5.

(not observed cells —expected cells!)

Formula

Degrees of Freedom (df)

• For all chi-square tests use the following: • df = (r – 1)(c – 1) • r is the row and c is the column

Calculator steps

TI-83+ Calculator Put your observed counts in L1 and Expected in L2 Then hit 2 nd and Stat Tyoe the observed in L1 and expected in L2. Then click on the L3 heading and type the formula(then click enter), then quit out to the main screen Find the sum of L3, your answer is the chi-square statistic

Calculator steps

After your get the sum you need to obtain the p-value X 2 , UB , df This will give you the p-value

Calculator steps

TI-84 calculator does most of the work for you Make sure you have typed your observed counts in L1 and expected counts in L2 5

Demographics.

• Rancho is approximately 53.6% Hispanic, 29.2% Asian, 12.7% white and 4.5% other.

(data as of 2012-2013 school year)

• Does Mr. Pines ’ AP stats classes reflect this diversity? Run the appropriate test, verify your requirements, and write a conclusion.

2015

Demographics.

H o : The diversity in Mr. Pines classes is the same as Rancho ’s diversity H a : The diversity in Mr. Pines classes is different than Rancho ’s diversity

Assumptions

: We have an independent random sample of 159 students ethnicity. We can assume that there have been at least 1590 students in Mr. Pines classes. 100% of our expected cells are 5 or more.

Chi-square GOF test

X

2 = 12.38

P-value =.0062

df = 3

This p-value is low enough to reject Ho at the 1% level This is strong evidence to suggest that Mr. Pines class diversity may be different than Rancho ’s diversity.

Is there a difference……

• Do boys and girls prefer types of social media?

• Please choose your favorite of the three below.

One Hundred Fifty Seven students were surveyed………….

We took a sample of 83 girls and a sample of 74 boys.

Girls Boys Instagram

36 19

Snapchat

25 36

Twitter

22 19

What’s the difference?

Goodness of Fit…..how well does the observed data match the expected….2 rows or 2 columns • • Homogeneity…..More than one sample is taken with one categorical variable in mind

(2+ Samples, 1 category)

• • Independence/Association…..Only one sample is taken and there are two or more categories.

(1 sample, 2+ categories)

One Hundred Fifty Seven students were surveyed………….

Girls Boys Instagram

36 (29.076) 19 (25.924)

Snapchat

25 (32.248) 36 (28.752)

Twitter

22 (21.675) 19 (19.325)

Hypotheses for Chi-Square Test for Homogeneity

H o : There is no difference between gender and social media preference H a : There is a difference between gender and social media preference OR H o : The proportions of boys and girls who prefer each type of social media are the same H a : The proportions of boys and girls who prefer each type of social media are different

H o : There is no difference between gender and social media preference H a : There is a difference between gender and social media preference Assumptions

: We have two independent random samples of students social media preferences(83 girls and 74 boys). There are obviously more than 830 girls and 740 boys in the population sampled from who use social media. 100% of our expected cells are 5 or more.

Chi-square test of homogeneity

X

2 = 6.96

P-value =.0307

df = 2

This p-value is low enough to reject Ho at the 5% level This is evidence to suggest that there may be a difference between gender and social media preference.

Period 1

Period 3

Referrals vs Days of week

Monday

12

Tuesday

5

Wednesday

9

Thursday

4

Friday

15 The table shows the number of students referred for disciplinary reasons to the principals office, broken down by day of the week.

Are referrals related to the day of the week?

BIRTHDAYS.

• Are Mr. Pines students birth months distributed in proportion to the number of days in each month?

• We can run a chi-square GOF based on the # of days in each month n = 153

C/O 2015 per 1 only H 0: Mr. Pines students birthday months are in proportion to the number of days in each month.

H a : Mr. Pines students birthday months are different than the proportion of the number of days in each month.

We have an independent sample of 53 students birth months. All births is obviously more than 10x our sample. 100% of our expected counts are 5 or more.

X

2 = 9.81

Chi-square GOF Test This p-value is too high to reject H o P-value = .5475

df = 11 Based on this sample, there is NOT enough evidence to suggest that Mr. Pines students birthday months are different than the proportion of the number of days in each month.

11 12 12 17

O

9 7 12 11 14 16 20 12

E

n = 153

BIRTHDAYS

To figure out the expected we need to think about the number of days in each month.

Jan 31 July 31 Feb 28 Aug 31 Mar 31 Sep 30 Apr 30 Oct 31 May 31 Nov 30 June 30 Dec 31 Total = 365 days

C/O 2014 H 0: Births are uniformly distributed by the # of days in each month H a : Births are not uniformly distributed by the # of days in each month CONDITIONS df = 11 We have an independent sample of 129 students birth months. All births is obviously more than 10x our sample. 100% of our expected counts are 5 or more.

X

2 = 7.75

P = .7355

This p-value is too high to reject H o .

There is not enough evidence to suggest that births are not uniformly distributed by the # of days in each month.

O E

12 10.956

5 9.896

12 10.956

12 10.603

13 10.956

7 10.603

9 10.956

14 10.956

11 10.603

11 10.956

8 10.603

15 10.956

Chi-Square Test for Homogeneity

• Data is given in a 2-way table • Expected counts are found by using a matrix on your calculator or by multiplying the (ROW TOTAL)(COLUMN TOTAL)/GRAND TOTAL • Conditions and df are the same as GOF test

Is there a difference……

• Do boys and girls prefer different video game consoles?

• Please choose your favorite console out of the 3?

Hypotheses for Chi-Square Test for Homogeneity

Remember Ha is always means different!

H o : There is no difference between gender and video game console preference H a : There is a difference between gender and video game preference OR H o : The proportions of boys and girls who prefer each type of console are the same H a : The proportions of boys and girls who prefer each type of console are different

Why is this a Homogeneity Test?

• Two samples were taken separately • Boys console preference • Girls console preference • There is ONE category of interest.

Two-Way Table

2013 G E N D E R

Girls Boys Total VIDEO GAME CONSOLE PREFERENCE Wii Xbox 360 PS3 27 3 30 16 33 49 20 32 52 63 68 131

Two-Way Table

2014 G E N D E R

Girls Boys Total VIDEO GAME CONSOLE PREFERENCE Wii Xbox Playstation 19 6 25 10 21 31 32 34 66 61 61 122

Hit 2 nd Matrix, go to EDIT Quit to home screen, go to test menu Set the appropriate matrix size Should be already setup if you used A and B Enter observed counts in matrix

You need the expected counts….so go back to 2 nd matrix. Use NAMES and go down to Matrix B, calculator generates them after you run the test You will most likely have to scroll to the right to see all of the expected counts

REMEMBER!....EXPECTED COUNTS MUST BE ON YOUR PAPER!

College Students’ Drinking

In 1987, a random sample of undergraduate students at Rutgers University was sent a questionnaire that asked about their alcohol drinking habits. Here are the results displayed in a two-way table.

Live on Campus Total

Abstain Light Drinker Light Moderate Moderate Heavy Heavy Total 46 71 55 63 67 302

Live Off Campus(Not with Parents)

17 38 34 24 28 141

Live Off Campus with Parents

43 42 26 15 17 143 106 151 115 102 112 586

Chi-Square Test for Independence

There was one sample taken and then data was broken down into different categories.

When only one sample is taken we are doing a Chi-Square Test for Independence/Association

Hypotheses

This is a chi-square test for Independence/Association, you have a few options for writing the hypotheses

H o : There is no association between students ’ residence type and drinking habits.

H a There is an association between students ’ residence type and drinking habits.

OR

H o : Student drinking habits and residence type are independent H a Student drinking habits and residence type are not independent

Full Moon

Some people believe that a full moon elicits unusual behavior in people. The table shows the number of arrests made in a small town during the weeks of six full moons and six other randomly selected weeks in the same year. Is there evidence of a difference in the types of illegal activity that takes place.

This is a chi-square test for Homogeneity

Violent (murder,assault,rape, etc.) Property(burglary, vandalism, etc.) Drugs/Alcohol Domestic Abuse Other offenses

Full Moon

2 17 27 11 9

Not Full

3 21 19 14 6

Thanks To: Grace Montgomery

THANKS TO: Amy Nguyen

Testing M&M

s

• The Mars company has always claimed that the color distribution of their M&M ’ s follow a certain proportion as follows:

Brown

13%

Red

13%

Yellow

14%

Green

16%

Orange

20%

Blue

24% Check the M&M ’ s that were given to you. How many of each color do you have? We will run a Chi-Square GOF test to see if their claim is accurate.

Do not eat your M&M ’ s until we have all observed and expected counts completed!

Hypotheses

• Ho: My bag of M&M ’ s follow the same color distribution as the Mars company claim.

• Ha: My bag of M&M ’ s follows a different color distribution as the Mars company claim.

Assumptions/Conditi ons

• ___% of expected counts >5 • My bag of M&M ’ s can be considered an independent random sample of M&M ’ s

Colors

Claim % Expected Observed

M&M Combined Results

Brown

13% 728.78

606

Red

13% 728.78

586

Yellow

14% 784.84

803

Green

16% 896.96

1101

Orange

20% 1121.2

1335

Blue

24% 1345.44

1175

There were a total of 5606 M&M

s sampled.

We have a chi-square statistic of 157.85 which gives a P-value of 0.

Mr. Pines Poker Chips

• • • 44 white chips = 20pts 5 blue chips = 30pts 1 red chip = 50pts

Mr. Pines Poker Chips

• • • 44 white chips = 20pts 5 blue chips = 30pts 1 red chip = 50pts There have been 135 attempts at randomly choosing poker chips out of the bag. A White chip has been pulled 301 times, a Blue chip 47 times, and the Red chip 16 times. Has this followed the expected probabilities? Run a chi-square GOF Test.

Baseball Bats

• There have been some major bat changes for the 2011 season. Aluminum baseball bats have been regulated so that they meet certain safety standards. After 5 games this season, coach Pines has noticed significant reductions in power numbers such as 2B ’ s, 3B ’ s, and HR ’ s…..Of course he would like to test his hypothesis.

Baseball Bats

• Run a Chi-Squared two-way table test to see if there is an association between the power numbers and types of bats.

• Also run a 2-Prop. Z Test between the types of bats used.

• If these are done correctly, Z 2 = X 2

Hypotheses

• Ho: There is no association between type of bats and extra base hits • Ha: There is an association between type of bats and extra base hits

Assumptions/Conditi ons

• E---All expected counts > 5 • S----We have a random sample of 23 schools hitting stats for the first 5 games of the 2010 and 2011 baseball seasons • I----We can assume that all stats are independent of other teams stats

Observed and Expected Counts

Singles Extra Base Hits

2010(BESR Bats)

672 (695.26) 313 (289.74)

2011(BBCOR Bats)

703 (679.74) 260 (283.26)

X

2 = 5.35

P-value = .0207

• This p-value is low enough to reject at the 5% level.

• There is evidence to suggest that there may be an association between the types of bats and extra base hits

Tootsie Pop Wrappers

We are interested in whether or not the designs on the wrappings on Tootsie Roll Pops are independent of the flavor of the pop.

Hypotheses

This is a chi-square test for Independence/Association, you have a few options for writing the hypotheses

H o : There is no association between pop flavor and designs on the wrapper.

H a There is an association between pop flavor and designs on the wrapper.

OR

H o : Pop flavor and wrapper designs are independent.

H a Pop flavor and wrapper designs are not independent.