Exam 6: Causation and Experimental Design

Experiments have relatively low external validity.

Association, time order, and nonspuriousness are required to establish a causal effect.

After they had failed the writing exam that was required of all students before graduation, fifteen students were placed in a class in which they received additional tutoring. After completing this class, all fifteen students retook the exam earned higher scores. Their success was held up as an example that the class curriculum really works. What source of internal invalidity has not been considered?

Describe each of the five criteria for identifying a causal effect and provide an example of each. Which are best demonstrated through experiments? Why? Which do experiments not demonstrate well? How can experimental designs be altered to compensate? Is anything lost with quasi-experimental designs?

A spurious relationship between two variables is:

Describe the difference between randomization and matching. What are the advantages and disadvantages of each?

A procedure that assesses that treatment and measurement are consistent throughout an experiment is a(n):

Sherman and Berk's (1984) study looking at the impact of arrest in actual domestic violence incidents is an example of a:

A university conducts a series of surveys in which subjects are drawn from first year and graduating students and finds that graduating students demonstrate much higher rates of approval for the university (in terms of whether they think the classes were relevant, the instructors were competent, etc.). The university then claims that more experience with the university leads to higher approval. What source of internal invalidity has the university not considered?

Which of the following are requirements for identifying a causal effect?

Use the following to answer questions 16-18: In each set, match a concept from Group A to a definition or example from Group B. -Sources of Internal Invalidity in Experiments Match the measurement procedure in Group A to the definition in Group B. Group A 1. Pretesting 2. Selection bias 3. The History Effect 4. The Hawthorne Effect 5. Demoralization 6. The John Henry Effect 7. Regression 8. Mortality Group B a. Things can happen outside the experiment that change subjects' scores b. Competition between control and experimental groups can change scores c. Subjects' scores can change because they feel special for being chosen for an experiment d. Subjects will drop out of the experimental or control group at rates that change the nature of the comparison between groups e. Measurement before experimental treatment may change subjects' scores on posttest f. Inferences about group processes are drawn from individual level data g. Subjects in the control group may feel deprived of treatment and perform worse on posttests than expected h. Extreme scores used to select subjects will move back toward the average in subsequent testing i. Subjects in the control and experimental group are initially different

Two studies are conducted of the relationship between SAT scores and subsequent earnings. In Study A, students are the units of analysis and it is found that college students who had higher SAT scores have higher earnings in subsequent jobs. Write a statement about this finding that would be reductionist. Explain why the statement is reductionist. In Study B, colleges are the units of analysis and it is found that colleges in which the average student SAT score was higher also have, on average, higher earnings among their graduates. Write a statement about this finding that exemplifies an ecological fallacy. Explain what makes this statement an ecological fallacy.

Two types of before-and-after experimental designs in which multiple pretest and posttest observations are made of the same group are:

Compensatory rivalry exists between comparison groups in an experiment when:

In an experiment on obedience to authority, one subject tips another off that the shocks that he thinks he is administering to another subject are not real. This is an example of:

You want to conduct research that documents the effects of watching violent cartoons on violence during children's playtime. Set up the following types of experimental and quasi-experimental designs: true experiment, non-equivalent group, before-and-after, ex-post facto, and Solomon four group. Which design do you think would produce the most valid results? Explain why.

For researchers to ensure the results of their experiment worked for other groups, they decided to make several comparisons involving the same independent and dependent variables but with different groups, also known as:

The reason that ex post facto control group designs are not considered quasi-experimental is:

Subjects who can be recruited for a laboratory experiment, randomly assigned to a group, and kept under carefully controlled conditions for the duration of the study are the most likely to be representative of the larger population.

The Hawthorne Effect suggests that if the control group believes it is being left out of some special treatment, it will perform worse than expected on posttests.