Chapter 9 Inferences from Two Samples

Transcription

1 Chapter 9 Inferences from Two Samples 9-1 Review and Preview 9-2 Two Proportions 9-3 Two Means: Independent Samples 9-4 Two Dependent Samples (Matched Pairs) 9-5 Two Variances or Standard Deviations

2 Review In Chapters 7 and 8 we introduced methods of inferential statistics. In Chapter 7: confidence interval. In Chapter 8: hypothesis testing. Chapters 7 and 8 both involved methods for dealing with a sample from a single population.

3 Preview The objective: testing hypotheses to situations involving two sets of sample data instead of just one.

4 Chapter 9 Inferences from Two Samples 9-1 Review and Preview 9-2 Two Proportions 9-3 Two Means: Independent Samples 9-4 Two Dependent Samples (Matched Pairs) 9-5 Two Variances or Standard Deviations

5 Key Concept 1. testing a claim made about two population proportions 2. constructing a confidence interval estimate of the difference between the two population proportions. This section is based on proportions, but can also deal with probabilities or percentages.

6 Notation for Two Proportions For population 1, we let: p 1 n 1 x 1 pˆ qˆ 1 = population proportion = size of the sample = number of successes in the sample = x n 1 1 = (the sample proportion) pˆ The corresponding notations apply to which come from population 2. p, n, x pˆ and qˆ, 2 2 2, 2 2

7 Pooled Sample Proportion The pooled sample proportion is given by: p = x n + x n 1 2 q = 1 p

8 Requirements 1. We have proportions from two independent simple random samples. 2. For each of the two samples, the number of successes is at least 5 and the number of failures is at least 5. (so that the binomial distribution can be approximated by the normal distribution)

9 Test Statistic for Two Proportions For H : p = p z = ( pˆ pˆ ) ( p p ) pq n + pq n 1 2 where (assumed in the null hypothesis) p 1 p2 = 0

10 Confidence Interval Estimate of p 1 p 2 ( pˆ pˆ ) E < ( p p ) < ( pˆ pˆ ) + E with margin of error E pq ˆˆ = zα /2 + n pq ˆˆ n

11 Example Do people having different spending habits depending on the type of money they have? 89 undergraduates were randomly assigned to two groups and were given a choice of keeping the money or buying gum or mints. The claim is that money in large denominations is less likely to be spent relative to an equivalent amount in many smaller denominations. Let s test the claim at the 0.05 significance level.

12 Example Below are the sample data and summary statistics: Group 1 Group 2 Subjects Given $1 Bill Subjects Given 4 Quarters Spent the money x 1 = 12 x 2 = 27 Subjects in group n 1 = 46 n 2 = 43 pˆ p = 12 pˆ = = =

13 Requirement Check: Example 1.The 89 subjects were randomly assigned to two groups, so we consider these independent random samples. 2.The subjects given the $1 bill include 12 who spent it and 34 who did not. The subjects given the quarters include 27 who spent it and 16 who did not. All counts are above 5, so the requirements are all met.

14 Example Step 1: The claim that money in large denominations is less likely to be spent can be expressed as p 1 < p 2. Step 2: If p 1 < p 2 is false, then p 1 p 2. Step 3: The hypotheses can be written as: H : p = p H : p < p 1 1 2

15 Example Step 4: The significance level is α = Step 5: We will use the normal distribution to run the test with: pˆ 12 ˆ 27 1 = p2 = p = = q = 1 p = =

16 Example Step 6: Calculate the value of the test statistic: z = = ( pˆ pˆ ) ( p p ) pq n + pq n ( )( ) ( )( ) = 3.49

17 Example Step 6: This is a left-tailed test, so the P-value is the area to the left of the test statistic z = 3.49, or The critical value is also shown below.

18 Example Step 7: reject the null hypothesis. Step 8: There is sufficient evidence to support the claim that people with money in large denominations are less likely to spend relative to people with money in smaller denominations.

19 Example We can also construct a confidence interval to estimate the difference between the population proportions. Caution: A conclusion based on a confidence interval might be different from a conclusion based on a hypothesis test.

20 Example Construct a 90% confidence interval estimate of the difference between the two population proportions. What does the result suggest about our claim about people spending large denominations relative to spending small denominations?

21 Example E pq ˆˆ = z + α /2 n pq ˆˆ n = =

22 Example ( pˆ pˆ ) E < ( p p ) < ( pˆ pˆ ) + E < ( p p ) < < ( p p ) < The confidence interval limits do not include 0, implying that there is a significant difference between the two proportions. There does appear to be sufficient evidence to support the claim that money in large denominations is less likely to be spent relative to an equivalent amount in many smaller denominations.

23 Chapter 9 Inferences from Two Samples 9-1 Review and Preview 9-2 Two Proportions 9-3 Two Means: Independent Samples 9-4 Two Dependent Samples (Matched Pairs) 9-5 Two Variances or Standard Deviations

24 Key Concept Using sample data from two independent samples to test hypotheses made about two population means or to construct confidence interval estimates of the difference between two population means.

25 Key Concept Part 1: the standard deviations of the two populations are unknown and are not assumed to be equal. Part 2: two other situations: (1) The two population standard deviations are both known; (2) the two population standard deviations are unknown but are assumed to be equal (compare to part 1) Because σ is typically unknown in real situations, most attention should be given to the methods in Part 1.

26 Part 1: Independent Samples with σ 1 and σ 2 Unknown and Not Assumed Equal

27 Definitions Two samples are independent if the sample values selected from one population are not related to or somehow paired or matched with the sample values from the other population. Two samples are dependent if the sample values are paired. (such as before/after data), or each pair of sample values consists of matched pairs (such as husband/wife data)

28 Notation μ 1 σ 1 n 1 x 1 s 1 = population mean = population standard deviation = size of the first sample = sample mean = sample standard deviation Corresponding notations apply to population 2.

29 Requirements 1. σ 1 and σ 2 are unknown and no assumption is made about the equality of σ 1 and σ The two samples are independent. 3. Both samples are simple random samples. 4. Either or both of these conditions are satisfied: The two sample sizes are both large (over 30) or both samples come from populations having normal distributions.

30 Hypothesis Test for Two Means: Independent Samples t = ( x x ) ( µ µ ) s 2 s n + n 1 2 (where μ 1 μ 2 is often assumed to be 0)

31 Confidence Interval Estimate of μ 1 μ 2 Independent Samples ( x x ) E < ( µ µ ) < ( x x ) + E where E 2 s s = t + α /2 n n 1 2 df = smaller of n 1 1 or n 2 1.

32 Example Researchers conducted trials to investigate the effects of color on creativity. Subjects with a red background were asked to think of creative uses for a brick; other subjects with a blue background were given the same task. Responses were given by a panel of judges. Researchers make the claim that blue enhances performance on a creative task. Test the claim using a 0.01 significance level.

33 Example The data: Red Background Blue Background Sample size Sample mean score x = 3.39 Sample standard deviation n = 35 s = 0.97 x = 3.97 n = 36 s = 0.63

34 Example Requirement check: 1.The values of the two population standard deviations are unknown and assumed not equal. 2.The subject groups are independent. 3.The samples are simple random samples. 4.Both sample sizes exceed 30. The requirements are all satisfied.

35 Example Step 1: μ 1 < μ 2. Step 2: If the original claim is false, then μ 1 μ 2. Step 3: H 0 : μ 1 = μ 2 H 1 : μ 1 < μ 2 (claim, left tail) Step 4: The significance level is α = 0.01 Step 5: we have two independent samples and we are testing a claim about two population means, σ is unknown, we use a t distribution.

36 Example Step 6: Calculate the test statistic t = ( x x ) ( µ µ ) s 2 s n + n 1 2 ( ) 0 = =

37 Example Step 6: Because we are using a t distribution, the critical value of t = is found from Table A-3. We use 34 degrees of freedom.

38 Example Step 7: reject the null hypothesis μ 1 μ 2 = 0. P-Value Method: P-value = reject the null hypothesis. Step 8. There is sufficient evidence to support the claim that the red background group has a lower mean creativity score than the blue background group.

39 Example Using the data from this color creativity example, construct a 98% confidence interval estimate for the difference between the mean creativity score for those with a red background and the mean creativity score for those with a blue background.

40 Example E s1 s = tα /2 + = = n n x = 3.39 and x = ( x x ) E < ( µ µ ) < ( x x ) + E < ( µ µ ) <

41 Example We are 98% confident that the limits 1.05 and 0.11 actually do contain the difference between the two population means. Because those limits do not include 0, our interval suggests that there is a significant difference between the two means.

42 Part 2: Alternative Methods These methods are rarely used in practice because the underlying assumptions are usually not met. 1.The two population standard deviations are both known the test statistic will be a z instead of a t and use the standard normal model. 2.The two population standard deviations are unknown but assumed to be equal pool the sample variances.

43 Hypothesis Test for Two Means: Independent Samples with σ 1 and σ 2 Known The test statistic will be: z = ( x x ) ( µ µ ) σ n σ + n

44 Confidence Interval Estimate for μ 1 μ 2 with σ 1 and σ 2 Known ( x x ) E < ( µ µ ) < ( x x ) + E E σ 2 σ = z + α /2 n n 1 2

45 Inference About Two Means, Assuming σ 1 = σ 2 The test statistic will be t = ( x x ) ( µ µ ) s n s 2 2 p p + n 1 2 where s ( n 1) s + ( n 1) s p = ( n1 1) + ( n2 1) df = n + n 2 1 2

46 Confidence Interval Estimate for μ 1 μ 2 Assuming σ 1 = σ 2 ( x x ) E < ( µ µ ) < ( x x ) + E E 2 s 2 p p s = t + α /2 n n 1 2 df = n + n 2 1 2

47 Methods for Inferences About Two Independent Means

48 Chapter 9 Inferences from Two Samples 9-1 Review and Preview 9-2 Two Proportions 9-3 Two Means: Independent Samples 9-4 Two Dependent Samples (Matched Pairs) 9-5 Two Variances or Standard Deviations

49 Key Concept Develop methods for testing hypotheses and constructing confidence intervals involving the mean of the differences of the values from two dependent populations.

50 Dependent Samples Example Matched pairs of heights of U.S. presidents and heights of their main opponents. Since there is a relationship as a basis for matching the pairs of data, this data consists of dependent samples. Height (cm) of President Height (cm) of Main Opponent

51 Good Experimental Design Using dependent samples with paired data is generally better than using two independent samples. The advantage of using matched pairs is that we reduce extraneous variation, which could occur if each experimental unit were treated independently.

52 Notation for Dependent Samples d = individual difference between the two values of a single matched pair µ d = mean value of the differences d for the population of all matched pairs of data d = mean value of the differences d for the paired sample data s d = standard deviation of the differences d for the paired sample data n = number of pairs of sample data

53 Requirements 1. The sample data are dependent. 2. The samples are simple random samples. 3. Either or both of these conditions is satisfied: The number of pairs of sample data is large ( n > 30) or the pairs of values have differences that are from a population having a distribution that is approximately normal.

54 Hypothesis Test Statistic for Matched Pairs t = d µ d s d n where degrees of freedom = n 1

55 Confidence Intervals for Matched Pairs d E < µ < d + E d E = t α /2 s d n Critical values of t α/2 : Use Table A-3 with df = n 1 degrees of freedom.

56 Example At α = 0.05, use the data below to test the claim that for the population of heights of presidents and their main opponents, the differences have a mean greater than 0 cm (so presidents tend to be taller than their opponents). Height (cm) of President Height (cm) of Main Opponent Difference d

57 Example Requirement Check: 1. The samples are dependent because the values are paired. 2. The pairs of data are randomly selected. 3. The number of data points is 5, so normality should be checked (and it is assumed the condition is met).

58 Example Step 1: The claim is that µ d > 0 cm. Step 2: If the original claim is not true, we have µ d 0 cm. Step 3: The hypotheses can be written as: H H 0 0 : µ = 0 cm d : µ > 0 cm d

59 Example Step 4: The significance level is α = Step 5: We use the Student t distribution. The summary statistics are: d = 3.2 s = 11.4

60 Example Step 6: Determine the value of the test statistic: t d µ = d = = s 11.4 d n 5 with df = 5 1 = 4

61 Example Step 6: Using technology, the P-value is Using the critical value method:

62 Example Step 7: Fail to reject the H 0. Conclusion: There is not sufficient evidence to support the claim that for the population of heights of presidents and their main opponent, the differences have a mean greater than 0 cm. In other words, presidents do not appear to be taller than their opponents.

63 Example Confidence Interval: Support the conclusions with a 90% confidence interval estimate for µ d. E = t s d α / 2 = = n d E < µ < d + E < µ < < µ < 14.1 d d d

64 Example The confidence interval does contain the value of 0 cm, so it is very possible that the mean of the differences is equal to 0 cm, indicating that there is no significant difference between the heights.

65 Chapter 9 Inferences from Two Samples 9-1 Review and Preview 9-2 Two Proportions 9-3 Two Means: Independent Samples 9-4 Two Dependent Samples (Matched Pairs) 9-5 Two Variances or Standard Deviations

66 Key Concept This section presents the F test for comparing two population variances (or standard deviations). Note that the F test is very sensitive to departures from normal distributions.

67 Part 1 F Test as a Hypothesis Test with Two Variances or Standard Deviations

68 Notation for Hypothesis Tests with Two Variances or Standard Deviations 2 s 1 n 1 2 σ 1 = larger of two sample variances = size of the sample with the larger variance = variance of the population from which the sample with the larger variance is drawn s, n, and σ are used for the other sample and population

69 Requirements 1. The two populations are independent. 2. The two samples are simple random samples. 3. The two populations are each normally distributed

70 Test Statistic for Hypothesis Tests with Two Variances F = s 2 1 s s 1 Where is the larger of the two sample variances P-values: Automatically provided by technology or estimated using Table A-5. Critical values: Using Table A-5, we obtain critical F values that are determined by the following three values: 1. The significance level α. 2. Numerator degrees of freedom = n Denominator degrees of freedom = n 2 1

71 Properties of the F Distribution The F distribution is not symmetric. Values of the F distribution cannot be negative. The exact shape of the F distribution depends on the two different degrees of freedom.

72 Finding Critical F Values

73 Properties of the F Distribution If the two populations do have equal variances, then F = s 2 1 s 2 2 s 2 1 s 2 2 will be close to 1 because and are close in value.

74 Properties of the F Distribution Consequently, a value of F near 1 will be evidence in favor of the conclusion that: σ 2 = σ But a large value of F will be evidence against the conclusion of equality of the population variances.

75 Example Subjects with a red background were asked to think of creative uses for a brick, while other subjects were given the same task with a blue background. Creativity score responses were scored by a panel of judges. Test the claim that those tested with a red background have creativity scores with a standard deviation equal to the standard deviation for those tested with a blue background. Use a 0.05 significance level. Creativity Score n s Red Background Blue Background x

76 Requirement Check: Example 1. The two populations are independent of each other. 2. simple random samples. 3. We have no knowledge of the distribution of the data, so we proceed assuming the samples are from normally distributed populations, but we note that the validity of the results depends on that assumption.

77 Example Step 1: The claim of equal standard deviations can be expressed as: σ = σ Step 2: If the original claim is false, then: σ σ Step 3: We can write the hypotheses as: H H : σ : σ = σ σ

78 Step 4: The significance level is α = Step 5: Because this test involves two population variances, we use the F distribution. Step 6: The test statistic is Example s 0.97 F = = = s The degrees of freedom for the numerator are n 1 1 = 34. The degrees of freedom for the denominator are n 2 1 = 35.

79 Step 6: Example From Table A-5, the critical F value is between and Technology can be used to find the exact critical F value of The P-value can be determined using technology: P-value =

80 Example Step 7: Since the P-value = is below the significance level of 0.05, we can reject the null hypothesis. Also, the figure below shows the test statistic is in the critical region, so reject.

81 Example There is sufficient evidence to warrant rejection of the claim that the two standard deviations are equal. The variation among creativity scores for those with a red background appears to be different from the variation among creativity scores for those with a blue background.

82 Part 2 Alternative Methods Two Methods that are not so Sensitive to Departures from Normality.