8b' Causation

1
8b. Causation Control

• Stanovich, Chapter 6

2
Review from Friday

• Correlation does NOT imply causation
• The directionality problem (what causes what?)
• E.g., do loving parent produce better adjusted
  children, or are better adjusted children just
  easier to love?
• 3rd variable problem (produces spurious
  correlations)
• E.g., toaster birth controlSES
• SES is a frequent 3rd variable

3
(No Transcript)
4
Freakonomics

• Reference
• Levitt, S., Dubner, S. (2005). Freakonomics A
  rogue economist explores the hidden side of
  everything. Morrow.
• www.freakonomics.com
• I had read the intro chapter where the authors
  cite the apparent observation that crime rates in
  the US declined about 20 years after Roe Vs. Wade

5

• It would appear that there is a negative
  correlation between abortion rate and crime rate
• The authors make an argument that legal abortion
  caused the crime rate to go down
• I raised this example as one that illustrates
  the temptation to infer causation in error.

6

• After each of my classes, students indicated that
  in a later chapter, the authors went into greater
  detail, citing other studies to support their
  position, and address alternative explanations.
• I havent read that chapter.
• To the extent that one is able rule out
  alternative hypotheses, one can be more confident
  that an observed correlation is not spurious.

7
Enter William Bennett

• Former Education Secretary and Drug Czar in US
  federal govt
• Now a political commentator/radio talk-show host
• News coverage over the weekend
• Hypothesized that aborting black babies in the US
  would lower the national crime rate
• Described this as ridiculous, impossible and
  morally reprehensible
• http//www.bennettmornings.com/

8
What do you think?

• controversy and condemnationjustified?
• Was there anything wrong with his thought
  experiment?
• Is this just spin, hype, or American politics as
  usual?
• Does this raise ethical questions about the
  report and presentation of research?

9
Bennetts problem

• He appeared to endorse the notion of a causal
  relationship between being black and being a
  criminal.
• He didnt state that the apparent correlation
  between those two factors was spurious, due to
  3rd variable (e.g., poverty)
• Must control for other variables before one can
  talk in terms of cause and effect

10
Selection Bias

• (p. 76 of text) Relationships between certain
  subject and environmental variables that may
  arise when people with different biological,
  behavioural, and psychological characteristics
  select different types of environments
• e.g.,
• Snow and cholera
• those that present to therapy may have more
  serious problems
• People who participate in psychology experiments
  are usually taking psychology classes and are
  more interested in human behaviour

11
Random Assignment

• Ensures that all conditions are roughly equal on
  all variables, because as sample size increases
  chance factors are balanced out
• What if you cant randomly assign?

12
Clever Hans

• WE NEED TO HAVE CONTROL (comparison) GROUPS TO
  RULE OUT ALTERNATIVE HYPOTHESES
• Another Example
• Facilitated communication
• Long words more difficult to recognize than short
  words

13
Problems We Need to Watch Out For

• Selection bias
• to avoid, use random assignment (if possible)
• Otherwise must accept as limitation of your study
  (correlational)

14

• Placebo Effect
• To rule out, use a control group
• Experimenter Bias when the researcher's
  preconceived ideas in some way influence the
  participants behaviour and/or the interpretation
  of experimental results
• To eliminate experimental bias experimenters use
  Double-blind Technique

15

• Demand Characteristics cues that participants
  pick up about the hypotheses of an experiment or
  how they are supposed to behave
• Think back to Ethics talk--deception