Chapter 15

Question 1

dependent samples design

Answer 1

• Dependent samples design: measurements in one sample are related to measurements in another sample
• ex. repeated measures, matched subjects, and matched pairs design

Question 2

repeated measures vs. matched subjects vs. matched pairs design

Answer 2

• Repeated measures design: same participants are tested in multiple conditions
• Matched subjects design: subjects are paired by experimenter
• Matched pairs design: subjects are naturally paired

Question 3

advantage of t-test for dependent samples

Answer 3

• higher positive r value -> smaller standard error -> bigger t -> bigger power
• Doing this removed a chunk of equation, which often outweighs the disadvantage of higher tcrit
• If you pick a zero correlation variable, it won’t work as well

Question 4

rho

Answer 4

correlation between pairs of observations in a population

Question 5

hedge’s g

Answer 5

• effect size; difference between sample means and estimated population stdvs
• .2 = small, .5 = medium, .8 = large

Question 6

r

Answer 6

correlational measure of effect size (partial correlation, NOT point-biserial)

Question 7

how to increase power?

Answer 7

increase n and pair observations to reduce standard error

Question 8

assumptions when testing hypotheses about different dependent means

Answer 8

• Assumption of homogeneity of variance NOT required
• Assumes sample is random and with replacement
• Normally distributed (if n >25, sampling distribution should be normally distributed)

Question 9

problems when testing hypotheses about different dependent means

Answer 9

• Must use random sampling
• Order effects: when exposure to first condition changes response to second condition -> increases standard error
• Non-random assignment: creates an order effect -> need a comparison group
• Sampling variation: correlation in sample may differ from correlation in population -> should only match subjects on variables that we know will have a significant influence on each other

Question 10

why doesn’t the homogeneity of variance assumption need to be tested?

Answer 10

Only needs to be tested when sample size is different, and n will always be the same in dependent samples