Chapter 11: Confounding and Obscuring Variables

Question 1

What is a design confound?

Answer 1

Alternative explanation due to poor experiment design

Question 2

What is a selection effect?

Answer 2

Conditions have systematically different subjects

Question 3

What is an order effect?

Answer 3

Only a problem in within-groups design, caused by order of levels presented

Question 4

What is a one-group pretest/posttest design?

Answer 4

1 group where a pre/posttest is performed with no comparison groups

Question 5

What is a maturation threat or spontaneous remission and how do you prevent it?

Answer 5

Improvement without known reasons, prevention by adding control group

Question 6

What is a history threat and how do you prevent it?

Answer 6

External factor systematically affects most members, such as seasonal changes. Prevention by adding control group

Question 7

What is regression to the mean and how do you prevent it?

Answer 7

Extreme scores are often less extreme when measured another time. This problem only occurs in groups that are measured twice and a group has an extreme score at pretest.
Prevention by adding comparison group

Question 8

What is an attrition threat and how do you prevent it?

Answer 8

It becomes a problem when the dropout is systematic
Prevention by removing participants’ scores from pretest average too

Question 9

What is a testing threat and how do you prevent it?

Answer 9

The behavior in participants change as a result of taking a test more than once. They can become more practiced or more tired etc.
Prevention by adding comparison group and by making it a post-test only design

Question 10

What is an instrumentation threat and how do you prevent it?

Answer 10

Changes in measuring instrument over time due to e.g. observers and different pre/posttests.
Prevention by doing post-test only and otherwise making sure pre/posttest are equivalent or counterbalancing pre/posttest

Question 11

What are four internal validity threats in any study?

Answer 11

1) Observer bias + effects
2) Demand characteristics
3) Non-specific threats
4) Placebo effects

Question 12

How can you prevent observer bias/effect and demand characteristics?

Answer 12

Doing a double-blind study where participants nor researcher knows who is in which condition
If that is hard, you can do a masked design, where participants know in which group they are, but the researchers don’t

Question 13

What is a non-specific threat and how can you prevent it?

Answer 13

If the effect is not due to treatment, but due to expectation to be treated
Prevention by doing a mixed factorial design with one waiting list group and the other group for treatment

Question 14

What are placebo effects and how can you prevent it?

Answer 14

People only improve, because they think they receive valid treatment.
Prevention: double-blind mixed facorial placebo control design (basically just add a placebo group to mixed factorial with waiting list design)

Question 15

What is a null effect?

Answer 15

If a confidence interval overlaps 0 or overlap completely between groups. There will always be a possibility that there is no effect

Question 16

What are the 4 problems with not enough between-groups differences?

Answer 16

1) Weak manipulations
2) Insensitive measures
3) Ceiling/floor effects
4) Reverse design confounds (unsystematic variability in each group)

Question 17

What are floor and ceiling effects?

Answer 17

Floor: test was too hard, so scores end up together at the low end
Ceiling: test was too easy, so scores end up together at the high end

Question 18

How can you prevent weak manipulations, insensitive measures and floor/ceiling effects?

Answer 18

By doing a manipulation check

Question 19

How can you prevent reverse design confounds?

Answer 19

Use reliable and precise tools, measure more people, change design into within-groups (if problems with individual differences), try controlling situational distractions

Question 20

Power is the opposite of obscuring. What is power?

Answer 20

Power is the likelihood a study will return an accurate result when de independent variable has an effect

Question 21

What is a reverse design confound?

Answer 21

There is too much unsystematic variability within each group and that obscures the differences between groups.