Deck 18: Repeated-Measures Analysis of Variance

Question 1

If, in the example of a headache study used in the text, all subjects had been studied over the same period of time, differences that might be caused by stressful times (such as Christmas) would

A) contaminate the results.
B) add to the usefulness of the results.
C) improve the generalizability of the study.
D) reduce the variability in the scores.

Answer 1

contaminate the results.

Question 2

All other things equal, the MS _error in a repeated-measures design is _______ than the corresponding MS _error in a between-subjects design.

A) smaller.
B) larger
C) the same size.
D) We have no way to predict this.

Answer 2

smaller.

Question 3

If we have a repeated-measures design with subjects receiving four levels of a treatment, we assume that

A) the correlations among the levels will be zero.
B) the correlations among the levels will be 1.0
C) the correlations among the levels will be random.
D) the correlations among the levels will all be about the same.

Answer 3

the correlations among the levels will all be about the same.

Question 4

The next few questions are based on the following summary table.

We don't have an F test on Subjects. What harm does that do?

A) It diminishes the value of the study.
B) It distorts the meaning of a significant trials effect.
C) It alters SS _total.
D) It does no particular harm.

Question 5

The next few questions are based on the following summary table.

If we wanted to run a set of multiple comparisons on the data analyzed in the summary table above, we could use

A) Fisher's LSD test.
B) the Repeated-Measures MC test.
C) the Bonferroni t test.
D) both a and c

Question 6

The next few questions are based on the following summary table.

The error term in this analysis could also be thought of as

A) the within cells term.
B) the between cells term.
C) the Subject × Trials interaction.
D) the variance.

Question 7

In running multiple comparisons in a repeated-measures design we can use procedures that we would use with independent groups designs because

A) the error term is corrected for any lack of independence in the data.
B) the error term is a conservative one.
C) the multiple comparison procedures we use have nothing to do with the question of independence or lack of it.
D) we create our own adjusted error term for multiple comparisons.

Question 8

If you are concerned that you have violated the assumptions behind a repeated-measures design, you can

A) limit your analysis to only those levels of the repeated measure that meet the assumption.
B) use an adjustment to your degrees of freedom.
C) pretend you didn't notice that there was a problem.
D) both a and b

Question 9

The assumptions behind the analysis of repeated-measures designs include

A) the assumption that all levels of the repeated measure are equally correlated with each other.
B) the assumption of normality.
C) the assumption of homogeneity of variance.
D) all of the above

Question 10

The next few questions are based on the following summary table.

If we compared Time 1 (baseline) against the next time (Time 2) and then against the last time (follow-up), we would run the Bonferroni at

A) α = .05.
B) α = .10.
C) α = .025.
D) α = .01.

Question 11

The next few questions are based on the following summary table.

How many subjects were involved in this study?

A) 14
B) 15
C) 74
D) 75

Question 12

The next few questions are based on the following summary table.

The results of this study tell us that

A) there is a change over trials.
B) there is no change over trials.
C) subjects are different.
D) There is too much noise in the system to draw any conclusions.

Question 13

The major advantage of repeated-measures designs is that

A) they allow you to use more subjects.
B) they allow you to remove subject differences from the error term.
C) they are easier to analyze.
D) they have a higher level of β.

Question 14

The next few questions are based on the following summary table.

The MS _error = 30.68 tells us that

A) the average variance in this study is 30.68 units.
B) the average variance within the set of scores for a given subject is 30.68 units.
C) the average difference due to time is 30.68 units.
D) We cannot tell from what is given.

Question 15

The next few questions are based on the following summary table.

How many trials were there in this experiment?

A) 2
B) 4
C) 5
D) 8

Question 16

All other things being equal, a repeated-measures design is _______ than the corresponding between-subjects design.

A) less powerful.
B) more powerful.
C) The two designs have comparable power.
D) We cannot predict this.

Question 17

The next few questions are based on the following summary table.

If we used a Bonferroni test to run multiple comparisons in the above example, the error term that we would use would be

A) the multiple comparison adjusted error term.
B) the MS _error from the table above.
C) the MS _subjects term.
D) MS _error/5.

Question 18

A repeated-measures analysis of variance differs from a one-way and a factorial design because

A) the measures in a repeated-measures design are correlated.
B) the measures in a standard factorial are independent.
C) the measures in a repeated-measures design are not independent from time to time.
D) all of the above

Question 19

For repeated-measures designs with one independent variable (Time),

A) subjects served under all levels of the independent variable.
B) subjects served under only some of the levels of the independent variable.
C) subjects served under only one level of the independent variable.
D) none of the above

Question 20

The next few questions are based on the following summary table.

How much has the error sums of squares been reduced over what it would have been in a comparable between-subjects design?

A) It has not been reduced at all.
B) 30.68 units
C) 723.5 units
D) 1687.3 units

Question 21

A Greenhouse and Geisser correction is a correction applied to

A) the mean.
B) the variance.
C) the degrees of freedom.
D) the interpretation.

Question 22

The text used an example in which the author rearranged the data points to look as if they came from a repeated-measures design. In real life we would not move our data points around so that we could analyze them as repeated measures. Why not?

A) It wouldn't work.
B) God would strike you dead.
C) It would be unethical to alter data in that way.
D) It is too awkward to do.

Question 23

In the printout of results for a repeated-measures analysis of variance, an F score for "mean" or "constant" sometimes appears. Why is this statistic often not interesting even if it is significant?

A) It shows differences between time sessions which are not important.
B) It is a randomly generated number.
C) It shows that the population mean is or is not equal to zero which is often of no interest.
D) It is redundant information given the F score for the time variable.

Question 24

A _______ design is one in which subjects are measured repeatedly over time.

A) between-subjects
B) factorial
C) repeated-measures
D) matched groups

Question 25

If a repeated-measures analysis of variance usually has an error term that is smaller that the error term in the corresponding between-subjects design, then we can assume that

A) repeated-measures designs have less power.
B) repeated-measures designs have greater power.
C) there are no differences due to power between the two kinds of designs.
D) neither design has very much power.

Question 26

In a typical learning experiment, a carry-over effect is

A) something to be avoided at all cost.
B) a necessary evil.
C) unlikely to be present.
D) what you are actually studying.

Question 27

Which of the following demonstrates the similarities of a repeated-measures analysis of variance for two trials and a t test for related means?

A) F = t ²
B) F ² = t
C) F = t
D) F = 2 t

Question 28

If both the Greenhouse and Geisser and the Huynh and Feldt corrections lead to significant results we should

A) not worry too much about our assumptions.
B) try to find a different way to analyze the data.
C) declare the uncorrected analysis to come have come to faulty conclusions.
D) decide that we should not believe our data.

Question 29

Counterbalancing is a technique to

A) lower the weight of subjects.
B) distribute carry-over effects evenly across the data.
C) increase the power of an experiment.
D) reduce the likelihood of reasonable conclusions.

Question 30

By shifting the data around the way the author did at the end of the repeated measures chapter, he was able to show that

A) the means became less different.
B) differences due to subjects were literally subtracted from the error term.
C) the F decreased.
D) all of the above

Question 31

In a repeated-measures ANOVA, tests to correct the degrees of freedom, such as Greenhouse-Gelsser and Huynh-Feldt, should be used if

A) you violate the assumptions of constant correlations.
B) you do not violate the assumptions of constant correlations.
C) you forget the assumptions of constant correlations.
D) you do not have an assumption of constant correlations.

Question 32

In a learning study using repeated measures, the correlation between early and later times will likely be low. Analyzing fewer levels of the independent variable would help to avoid violating the assumption of

A) normality.
B) homogeneity of variance.
C) constant correlations among pairs of levels of the repeated variables.
D) MS_error is an unbiased estimate of the magnitude of effect of the predictor variable in a regression analysis.

Question 33

The major disadvantage with repeated-measures designs is that they

A) require too many subjects.
B) are less powerful than between-subjects designs.
C) have a funny looking summary table.
D) are subject to the influence of carry-over effects.

Question 34

If the assumption of constant correlations in a repeated-measures ANOVA is violated, which of the following is true?

A) The F-value is incorrect.
B) The degrees of freedom should be adjusted.
C) The sample cannot be valid.
D) The manipulation checks failed.

Question 35

In an example in the text, an independent samples analysis of variance example from a previous chapter was converted to be used in a repeated-measures analysis of variance. Recalculating the F value with a repeated-measures analysis of variance yields an F value that is

A) less than the F value yielded by the independent measures ANOVA.
B) greater than the F value yielded by the independent measures ANOVA.
C) the same as the F value yielded by the independent measures ANOVA.
D) not predictably different from the F value yielded by the independent measures ANOVA.

Question 36

The typical way to control sequence effects is called

A) block randomization.
B) cross-sectional experimentation.
C) asymmetrical transfer.
D) counterbalancing.

Question 37

If any reasonable person would expect that with 4 trials the last trial is almost certain to be significantly different from the first, then Fisher's LSD test

A) has more protection against a high familywise error rate.
B) has a smaller degree of protection against a high familywise error rate.
C) will lead to a very high familywise error rate.
D) Error rates have nothing to do with the issue.

Question 38

A design in which each subject receives all levels of an independent variable is called a(n)

A) independent samples t-test.
B) repeated-measures design.
C) between-subjects design.
D) correlation.

Question 39

Some summary tables include a term labeled "mean" or "constant," with a corresponding F test. This tests the hypothesis that

A) the mean is equal to each of the population means.
B) the mean of the scores is equal to the population mean.
C) the population mean is 0.00.
D) subjects are all equal.

Question 40

A researcher wanted to see how watching movies influenced subjects' IQ scores. She gave IQ tests to subjects following each of two movies. Half of the subjects first saw Titanic followed by Schindler's List , while the other half first saw Schindler's List and then Titanic . Varying the movie order is an example of

A) counterbalancing.
B) random sampling.
C) selection bias.
D) practice effects.

Question 41

In a repeated-measures design, the error term is equivalent to the interaction between subjects and the repeated-measures factor.

Question 42

If reaction time data were collected from the same 10 people, 4 times, the df for time in a repeated-measures ANOVA based on that data would be 3.

Question 43

On the following computer output, the significance of F varies depending on which test you look at.

a. Why is this the case?
b. Which F value should be reported? Explain your answer.

Question 44

A researcher collected data from behaviorally disturbed youth to see if introducing a token economy would reduce their disruptive behavior. He collected 3 weeks of data at baseline, treatment, and withdrawal respectively. Identify three meaningful multiple comparisons you could calculate based on this data. Explain your answers.

Question 45

Give two examples in which you might use a repeated-measures design.

Question 46

Answer the following questions based on the summary table below.

a. How many people were in the sample?
b. How many times was the dependent variable measured?
c. Was there a difference in the dependent variable over time?   Explain.

Question 47

Give an example in which counterbalancing might be important for a repeated-measures design.

Question 48

Measuring the height of the same group of children every year for three years is an example of a between-subjects design.

Question 49

In a repeated-measures design, each subject receives all levels of at least one independent variable.

Question 50

If we ran a repeated-measures analysis of variance to track changes in patients' distorted thoughts over 6 weeks of therapy, we would most likely want to report the effect size in terms of

A) eta-squared.
B) omega-squared.
C) computed on the difference between adjacent trials.
D) computed on the difference between the initial trial and the last trial.

Question 51

Counter-balancing is an appropriate strategy to deal with practice or carry-over effects in repeated-measures designs.

Question 52

Explain why SS_subjects is removed from SS_error in repeated-measures designs.

Question 53

You want to run a study examining the effects of poverty on the development of antisocial behavior. You randomly select a large group of normal 12- year-old children and sort them into three groups on the basis of family income. You meet with them yearly until they are 25 years old, using a standard assessment of antisocial behavior. What test should you run to analyze this data?

A) independent samples ANOVA
B) one-way ANOVA
C) repeated-measures ANOVA
D) chi-square

Question 54

The Greenhouse-Geisser correction is used in repeated-measures ANOVAs when there is limited power due to restricted sample size.

Question 55

If the t for related means = 4, then the F for repeated measures based on the same data would = 16.

Question 56

Calculate and interpret F based on the following data.

Question 57

In a repeated-measures design, SS_error is calculated the same as it is in a between-subjects design.

Question 58

An assumption underlying repeated-measures ANOVAs is that pairs of levels of the repeated factor are uncorrelated.

Question 59

If reaction time data were collected from the same 10 people, 4 times, the total df in a repeated-measures ANOVA based on that data would be 40.

Question 60

Calculate and interpret F for the following example.

Question 61

A researcher examined reaction time in 12 people across 3 conditions: regular cola, caffeine free cola, and water. The overall F was significant, so she performed multiple comparisons to understand which conditions differed. Interpret the following multiple comparisons at the .05 level.
_cola = 2.43s,
_{caffeine free} = 2.52s,
_water = 2.53s.  t_{cola/caffeine free} = 2.80; t_cola/water = 2.17;  t _{water/caffeine free} = 0.38

Question 62

Given the following data from a repeated-measures design, what is the value for SS_error?
SS_subjects = 950, SS_time = 1500, SS_total = 3100