1. Correlation AND EXPERIMENTAL DESIGN
Chapter 6

2. Establishing causation
It appears that lung cancer is associated with smoking.
How do we know that both of these variables are not being affected by an unobserved third (lurking) variable?
What if there is a genetic predisposition that causes people to both get lung cancer and become addicted to smoking, but the smoking itself doesn’t CAUSE lung cancer?
The association is strong.
The association is consistent.
Higher doses are associated with stronger responses.
Alleged cause precedes the effect.
The alleged cause is plausible.
THERE IS NO SUBSTITUTE FOR AN EXPERIMENT!!!
We can evaluate the association using the following criteria:

3. 64% of American’s answered “Yes” . 38% replied “No”.
In a Gallup poll, surveyors asked, “Do you believe correlation implies causation?’”
64% of American’s answered “Yes” .
38% replied “No”.
The other 8% were undecided.

4. Causal Explanation
Cause: An explanation for some characteristic, attitude, or behavior of groups, individuals, or other entities
Causal effect: The finding that change in one variable leads to change in another variable, other things being equal.

5. Five Criteria for Identifying a Causal Effect
3 required
Association: Empirical (observed) correlation between independent and dependent variables (must vary together)
2. Time Order: Independent variable comes before dependent variable
3. Nonspuriousness: Relationship between independent and dependent variable not due to third variable

6. Criteria for Identifying a Causal Effect
These two strengthen the causal argument
4. Mechanism: Process that creates a connection between variation in an independent variable and variation in dependent variable
5. Context: Scientific explanation that includes a sequence of events that lead to particular outcome for a specific individual
Can not be used to explain general ideas, places, events, or populations

7. Correlation vs Causation
Topic 5, Lecture 3
Correlation vs Causation
Correlation tells us two variables are related
Types of relationship reflected in correlation:
X causes Y or Y causes X (causal relationship)
X and Y are caused by a third variable Z (spurious relationship)

8. Correlation vs. Causation Example
‘‘The correlation between workers’ education levels and wages is strongly positive”
Does this mean education “causes” higher wages?
We don’t know for sure !
Correlation tells us two variables are related BUT does not tell us why

9. Correlation vs. Causation
Possibility 1
Education improves skills & skilled workers get better paying jobs
Education causes wages to 
Possibility 2
Individuals are born with quality A, which is relevant for success in education and on the job
Quality A (NOT education) causes wages to 

10. Kids’ TV Habits Tied to Lower IQ Scores
IQ scores and TV time
r = -.54

11. Eating Pizza ‘Cuts Cancer Risk’ Pizza consumption and cancer rate

12. Reading Fights Cavities
Number of cavities in elementary school children & their vocabulary size
r = -.67

13. Stop Global Warming: Become a Pirate
Average global temperature and number of pirates
r = -.93

14. 4. Implying causation where only correlation exists

15. There are two relationships which can be mistaken for causation:
A strong relationship between two variables does not always mean that changes in one variable causes changes in the other.
The relationship between two variables is often influenced by other variables which are lurking in the background.
There are two relationships which can be mistaken for causation:
Common response
Confounding

16. Common response Confounding
Possibility that a change in a lurking variable is causing changes in both explanatory variable and response variable
Confounding
Possibility that either the change in explanatory variable is causing changes in the response variable OR
That change in a lurking variable is causing changes in the response variable.

17. 1. Common Response:
Both X and Y respond to changes in some unobserved variable, Z.

18. 2. Confounding
The effect of X on Y is indistinguishable from the effects of other explanatory variables on Y.
Example of confounding:
The “placebo effect”

19. When can we imply causation?
When controlled experiments are performed.
Unless Data Have Been Gathered By Experimental Means and Confounding Variables Have Been Eliminated, Correlation Never Implies Causation

20. Strongest for demonstrating causality
EXPERIMENTS
Strongest for demonstrating causality
Asch Experiment
Quasi-experimental designs
Looks like experimental design but lacks -- random assignment
Attraction and Scary Bridge

21. Most powerful design for testing causal hypotheses
Experiments
Most powerful design for testing causal hypotheses
Experiments establish:
Association
Time order
Non-spuriousness

22. Experimental design

23. True Experiments Two comparison groups to establish association
Experimental Group:
Treatment or experimental manipulation
Control group:
No treatment

24. Time order
Variation must be collected before assessment to establish time order
Post-test: Measurement of the DV in both groups after the experimental group has received treatment
Pre-test: Measurement of the DV prior to experimental intervention
True experiment doesn’t need a pre-test
Random assignment assumes groups will initially be similar

25. Non-spuriousness Random assignment (randomization):
Of subjects into experimental and control groups
Establishes non-spuriousness
Not random sampling
Randomization has no effect on generalizability

26. Matching Can only be done on a few characteristics
Assignment of subject pairs into experimental and control groups
Based on similarity (e.g., gender, age)
Individuals (in pairs) randomly assigned to each group
Can only be done on a few characteristics
May not distribute characteristics between the two groups

27. Quasi-Experimental Designs
Establish time order & association
May be better at establishing context
Cannot establish non-spuriousness
Comparison groups not randomly assigned

28. Internal validity in experiments
Confidence in cause and effect relationship
Key question in any experiment is:
“Could there be an alternative cause, or causes, that explains the observations and results?”

29. Generalization: External validity
Whether results from small sample group, in a laboratory, can be extended to make predictions about entire population

30. Problems with experiments
Threats to validity in experiments
True experiments have high internal but low external validity
Quasi-experiments have higher external but lower internal validity

31. Non-comparable groups
Experimental and Control groups are not comparable
Selection bias: subjects in experimental and control groups are initially different
Mortality/Differential attrition: groups become different because subjects are more likely to drop out of one of the groups for some reason
Instrument decay: Measurement instrument wears out or researchers get tired or bored, producing different results for cases later in the research than earlier

32. Endogenous change
Natural developments in subjects, independent treatment, account for some or all of change between pre- and post-test scores
Generally, eliminated by use of control group
Changes same for both groups.
Testing: Pre-test can influence post-test scores
Maturation: Changes may be caused by aging of subjects
Regression to the mean: When subjects are selected based on extreme scores
In future testing: Regress back to average

33. Things happen outside experiment may change subjects’ scores
The History effect
Things happen outside experiment may change subjects’ scores

34. Control and experimental groups affect one another
Contamination
Control and experimental groups affect one another

35. Demoralization: Contamination
The control group may feel left out and perform worse than expected

36. Contamination Compensatory Rivalry (The John Henry Effect):
When groups know being compared
May increase efforts to be more competitive

37. Treatment Misidentification
Expectancies of Experimental Staff:
Staff actions and attitudes change the behavior of subjects (i.e., a self-fulfilling prophecy)
Resolved by double-blind designs
Neither the subject nor the staff knows who’s getting the treatment and who’s not

38. Treatment Misidentification
Placebo Effect:
Subjects change because of expectations of change, not because of treatment itself
Hawthorne Effect:
Participation in study may change behavior simply because subjects feel special for being in the study

39. generalizability of experiments
More artificial experimental arrangements
Greater problem of sample generalizability
Subjects are not randomly drawn from population

40. generalizability of experiments
Field experiments:
Conduct experiments in natural settings
Increases ability to generalize.
Random assignment is critical