# Clarifying the Concepts 11-1. What is an ANOVA?

Answer: An ANOVA is a hypothesis test with at least one nominal independent variable (with at least three total groups) and a scale dependent variable. 11-3. The F statistic is a ratio of between-groups variance and within-groups variance. What are these two types of variance? Answer: Between-groups variance is an estimate of the population variance based on the differences among the means; within groups variance is an estimate of the population variance based on the differences within each of the three (or more) sample distributions. 11-5. What are the three assumptions for a between-groups ANOVA? Answer: The three assumptions are that the participants were randomly selected, the underlying populations are normally distributed, and the underlying variances of the different conditions are similar, or homoscedastic. 11-7. Why is the F statistic always positive? Answer: The F statistic is calculated as the ratio of two variances. Variability, and the variance measure of it, is always positive—it always exists. Variance is calculated as the sum of squared deviations, and squaring both positive and negative values makes them positive. 11-9. Explain the concept of sum of squares Answer: With sums of squares, we add up all the squared values. Deviations from the mean always sum to 0. By squaring these deviations, we can sum them and they will not sum to 0. Sums of squares are measures of variability of scores from the mean. 11-11. What is the grand mean? Answer: The grand mean is the mean of every score in a study, regardless of which sample the score came from. 11-13. We typically measure effect size with _________ for a z test or a t test and with _________ for an ANOVA. Answer: Cohen’s d; R2 11-15. What does post-hoc mean, and when are these tests needed with ANOVA? Answer: Post-hoc means "after this." These tests are needed when our ANOVA is significant and we want to discover where the significant differences exist between our groups.

11-17. Find the error in the statistics language in each of the following statements about z, t, or F distributions or their related tests. Explain why it is incorrect and provide the correct word. a. The professor reported the mean and standard error for the final exam in the statistics class. b. Before we can calculate a t statistic, we must know the population mean and the population standard deviation. c. The researcher calculated the parameters for her three samples so that she could calculate an F statistic and conduct an ANOVA. d. For her honors project, Evelyn calculated a z statistic so that she could compare a sample of students who had ingested caffeine and a sample of students who had not ingested caffeine on their video game performance mean scores. Answer: a. Standard error is wrong. The professor is reporting the spread for a distribution of scores, the standard deviation. b. t statistic is wrong. We do not use the population standard deviation to calculate a t statistic. The sentence should say z statistic instead. c. Parameters is wrong. Parameters are numbers that describe populations, not samples. The researcher calculated the statistics. d. z statistic is wrong. Evelyn is comparing two means; thus, she would have calculated a t statistic. 11-19. What are the four assumptions for a within-groups ANOVA? Answer: The four assumptions are that (1) the data are randomly selected, (2) the underlying population distributions are normal, (3) the variability is similar across groups, orhomoscedasticity, and (4) there are no order effects. 11-21. Explain the source of variability called ―subjects.‖ Answer: The "subjects" variability is noise in the data caused by each individual’s personal variability. It is calculated by comparing each person’s mean response across all levels of the independent variable with the grand mea n, the overall mean response across all levels of the independent variable. Calculating the Statistics 11-23. Calculate each type of degrees of freedom for the following data, assuming a between-groups design:

1970: 45, 211, 158, 74 1980: 92, 128, 382 1990: 273, 396, 178, 248, 374 a. dfbetween b. dfwithin c. dftotal

c.79 and within-groups variance is 2. for each of the following cases:a.05. Compute the missing values. Answer: 11-31. + dflast = (4 – 1) + (3 – 1) +(5 – 1) = 3 + 2 + 4 = 9 c. between-groups ANOVA source table is shown below. Between-groups variance is 2. dfbetween = Ngroups – 1 = 3 – 1 = 2 b.45. writing the ratio accurately.26. .05 is 4.20. Calculate a mean for each group and a grand mean for these data from Exercise 11-23. Using the F table and a p level of 0. Calculate the F statistic. . Within-groups variance is 41.83 and within-groups variance is 177. An incomplete one-way. dftotal = dfbetween + dfwithin = 2 + 9 = 11 11-25.24. Answer: 11-29. dfwithin = df1 + df2 + . Between-groups variance is 321. Answer: The critical value for a between-groups degrees of freedom of 2 and a within-groups degrees of freedom of 9 at a p level of 0. determine the critical value for the degrees of freedom determined in Exercise 11-23. 11-27. .Answer: a.60 and between-groups variance is 34.b.

374 Answer: 11-33. Total sum of squares is calculated here as SStotal = Σ(X -GM)2: . 211. Total sum of squares b. Using the data from Exercise 11-23. Within-groups sum of squares c. 382 1990: 273.1970: 45. Between-groups sum of squares Answer: (Note: The total sum of squares will not exactly equal the sum of the between-groups and within-groups sums of squares because of rounding error. calculate each of the following for a between-groups design: a. 178. 74 1980: 92. 396. 248. 158.) a. 128.

b. Within-groups sum of squares is calculated here as SSwithin = Σ(X – M)2: c. Between-groups sum of squares is calculated here as SSbetween = Σ(X – GM)2: .

we must calculate a weighted sample size. perform the simple division to complete an entire ANOVA source table for these data. Answer: 11-37. Remember that this F statistic was based on the data originally presented in Exercise 11-22 and that you worked with in Exercise 11-32. Calculate Tukey HSD values for the necessary comparisons following the F statistic you calculated in Exercise 11-34. . Answer: Because we have unequal sample sizes. Using all of your calculations from Exercises 11-23 and 11-33.11-35.

Calculate each type of degrees of freedom for the following data.Now we can compare our three groups. 11-39. assuming a within-groups design: .

dftotal = dfbetween + dfsubjects + dfwithin = 2 + 3 + 6 = 11. or dftotal = Ntotal − 1 Answer: a. dfbetween = Ngroups – 1 = 3 – 1 = 2 b.a. or we can calculate it as dftotal = Ntotal – 1 = 12 – 1 = 11 11-41. Answer: . dfsubjects = n − 1 c. dfwithin = (dfbetween)(dfsubjects) d. perform the simple division to complete an ANOVA source table for these data. dfsubjects = n – 1 = 4 – 1 = 3 c. dfwithin = (dfbetween)(dfsubjects) = (2)(3) = 6 d. Using all of your calculations in Exercises 11-39 and 11-40. dftotal = dfbetween + dfsubjects + dfwithin. dfbetween = Ngroups − 1 b.

students to identify their political viewpoint as most similar to that of the Republicans. b.c. Finally. This involves calculating a measure of variability among the three sample means—the religiosity scores of the Republicans. So it is likely that the distribution of sample means is normal and that this assumption was met.S. a measure that assesses how religious one is. Democrat. Democrats. The independent variable is political viewpoint. time on the job. The researchers wondered whether political orientation affected levels of religiosity. we would divide the between-groups variance by the within-groups variance. and high trust in supervisors. What is the third assumption for ANOVA? How could the researchers check to see if they had met this assumption? Be specific. 11-53. or length of time working with the supervisor. we would calculate the between-groups variance. The second assumption is that the population distribution is normal. this is essentially an average of the variability within each of the three samples. What are the populations and what are the samples? d. All three groups then completed a religiosity scale. Democrats. The researchers could assess whether this assumption was met by comparing the variance of each of the three treatment groups to ensure that the largest variance is no larger than two times the smallest variance. What is the dependent variable? c. First. So it is likely that the Yale researchers used a sample of participants from their local area. When there are at least 30 participants in a sample. c. and neither. The first assumption of ANOVA is that the samples are randomly selected from their populations. It is not possible to tell whether this assumption was met on thebasis of the study description. most similar to that of the Democrats. we introduced a study by Steele and Pinto (2006) that examined whether people’s level of trust in their direct supervisor was related to their level of agreement with a policy supported by that leader. a. and what are its levels? b. d. and all who categorize themselves as neither. moderate trust. What is the independent variable. this provides evidence that the means are different from each other. The study description indicates that the researchers were from Yale University and does not mention any techniques the researchers might have used to obtain participants from across the nation. If the variability among the means is much larger than the variability within each sample. b. there are more than 30 participants in the study. regardless of the specific religion with which a person identifies. Researchers asked 180 U. all Democrats. Answer: a. with the levels Republican. The dependent variable is religiosity. The populations are all Republicans. and people who say they are neither among the 180 students. or neither. It is unlikely that the researchers met this assumption. Let’s assume we used a scale that sorted employees into three groups: low trust. c. age. We have presented . Steele and Pinto found that the extent to which subordinates agreed with their supervisor was related to trust and showed no relation to gender. Then we would calculate the within-groups variance. the distribution of sample means will be approximately normal even if the underlying distribution of scores is not. Although we do not know the exact distribution of the population of scores. The samples are the Republicans. explain how you would calculate the F statistic. The third assumption is homoscedasticity—that our samples come from populations with equal variances. and others. Using this example. In Chapter 10. Answer: a. 11-55.

Step 5: dftotal = dfbetween + dfwithin = 2 + 8 = 10 GM = 21. . Step 4: The critical value for the F statistic based on a p level of 0. + dflast = (4 – 1) + (3 – 1) + (4 – 1) = 3 + 2 + 3 = 8 The comparison distribution will be the F distribution with 2 and 8 degrees of freedom.05 is 4.Step 3: dfbetween = Ngroups – 1 = 3 – 1 = 2 dfwithin = df1 + df2 + .727 Total sum of squares is calculated here as SStotal = Σ(X – GM)2: Within-groups sum of squares is calculated here as SSwithin = Σ(X – M)2: .46. .

Between-groups sum of squares is calculated here as SSbetween = Σ(X – GM)2: .

The mean level of agreement with a policy supported by a supervisor varies across level of trust in that supervisor. is beyond our cutoff of 4. Now we can compare our three groups. What did you learn? Answer: Because we have unequal sample sizes. we must calculate a weighted sample size. so we can reject the null hypothesis. 11-57. . our research design and data did not meet the three assumptions of this statistical test.Step 6: Our F statistic.46. Remember. 5.49. you conducted an ANOVA on data regarding employees’ trust in supervisors. Conduct a Tukey HSD test. In Exercise 11-55. so we should be careful in interpreting this finding.

The data presented were hours spent wearing the appliance per day. 16. there is a statistically significant difference between the mean level of agreement by employees with low trust in their supervisors compared to those with high trust. We obtained one q value (-4. complete the accompanying software output tables on pg 295. b. (Hint: Would we expect that the ―Sig. We can compare the actual p level to a cutoff p level such as 0. The table for the independent-samples t test and the table for the oneway betweengroups ANOVA were calculated using the identical fictional data. For the t test. In statistical software output. 13. 18 Junior high school: 8.‖ for the independent -samples t test be the same as or different from that for the one-way between-groups ANOVA?) . (Hint: Use the F statistic that you calculated in part (a). Based on your knowledge of the relation between the t and F distributions. what is the ―Sig. (Hint: The ―Mean Square‖ column includes the two estimates of variance used to calculate the F statistic.) b.‖? Explain how you determined this. What is the F statistic? Show your calculations.‖ refers to the actual p level of the statistic. Because the sample sizes here were so small and we did not meet the three assumptions of ANOVA. 18 a. Calculate the appropriate measure of effect size for this sample. medium. junior high school. Based on our calculations. and high school. a. Primary school: 16. In Exercise 11-54 we used a one-way between-groups ANOVA to explore patients’ likelihood of wearing orthodontic appliances.According to the q table.05 to decide whether to reject the null hypothesis. 14. Based on Cohen’s conventions. Why is it useful to have this information in addition to the results of a hypothesis test? Answer: a. 15. The researchers compared students in primary school. these preliminary findings would encourage additional research.) c. there’s a large effect. 12 High school: 20. 11-61. In fact. What is the t statistic? Show your calculations. is this a small.363) that exceeds this cutoff. It is useful to have this information because hypothesis testing tells us only whether year in school is a significant factor affecting how long patients wear their appliances. c. or large effect size? c.05 when you are comparing three groups and have a within-groups degrees of freedom of 8. 11-59. The appropriate measure of effect size is b. 13. the critical value is 4. we should be careful in making strong statements about this finding. ―Sig. According to Cohen’s conventions. R2 gives us an indication of how large this effect is —or how much of the variability in appliance wearing can be accounted for by year in school. this is a large effect size.04 for a p level of 0. Here.

Conduct all six steps of the hypothesis test.005. Here are some hypothetical data: a. Imagine a student decides to compare four different gums using five participants. .Answer: c. Let’s put these claims to a test. each participant had chewed all four types of gum. some commercials present the idea that the flavor lasts too long. Each randomly selected participant was asked to chew a different piece of gum each day for four days." for t is the same as that for the ANOVA. b. The order of the gums was randomly determined for each participant. The "Sig. 11-63. participants recorded the intensity of flavor from 1 (not intense) to 9 (very intense). Are any additional tests warranted? Explain your answer. In fact. 0. After two hours of chewing. such that at the end of the four days. because the F distribution reduces to the t distribution when we are dealing with two groups. Commercials for chewing gum make claims about how long the flavor will last. affecting sales and profit.

Step 1: Four different populations are being compared here: one for the people who chew gum 1. we can also calculate it as: dftotal = Ntotal – 1 = 20 – 1 = 19 We will use the F distribution with 3 and 12 degrees of freedom. within-groups ANOVA. The comparison distribution will be the F distribution.49. We do not know anything about the underlying population distributions. across the different gum types.Answer: a. Step 3: We determine the characteristics of the comparison distribution by computing our degrees of freedom. Step 4: The critical value for F with 3 and 12 degrees of freedom at a p level of 0.8 . SStotal = Σ(X – GM)2 = 111. and we will calculate a oneway. Research hypothesis: People do not report the same flavor intensity. We will proceed with caution. across the different gum types. dfbetween = Ngroups – 1 = 4 – 1 = 3 dfsubjects = n – 1 = 5 – 1 = 4 dfwithin = (dfbetween)(dfsubjects) = (3)(4) = 12 dftotal = dfbetween + dfsubjects + dfwithin = 3 + 4 + 12 = 19. on average.05 is 3. and we must be careful when assessing homoscedasticity with such a small sample. Step 5: Calculate the test statistic. on average. and one for each of the other three types of gum. Step 2: Null hypothesis: People report the same flavor intensity. Regarding our assumptions. we know that the participants were randomly selected and that the order of presentation of the gum was also randomly determined.

80 .SSbetween = Σ(X – GM)2 = 93.

3 .SSsubjects = Σ(Mparticipant – GM)2 = 2.

.

23. computing the effect size can give us an indication of just how large or important our finding is. Also. we would want to compute Tukey HSD statistics. or conduct some other post-hoc test. b. is well beyond our critical value of 3.Step 6: Our F statistic. so we would reject the null hypothesis and conclude that there are mean differences in the intensity of flavor. however. . without a post-hoc test.49. We cannot conclude which gums are different. to determine where the significant mean differences are within our data. because we computed a significant F statistic.90. Yes.