Casual Relationships

Question 1

What are the two kinds of spurious relationships and what characteristics tell them apart

Answer 1

Chance (false positive error): Two variables are correlated however it may be due to chance

Third variable (confounding error): Two variables are not correlated, may be a third variable

Question 2

What are some of the p-hacking techniques that might make a chance spurious relationship look significant? How might cherry picking results be used to the same effect?

Answer 2

Running participants in batches until p<.05

Subgroup analysis (effect is found in one gender, only one personality type etc)

Adding covariance (things to “control” for that make sense only after the fact, such as age, etc)

Cherry picking (may only pick out significant data, data that fits your narrative)

Question 3

Know Hume’s three requirements for causation and which is the most difficult and un-intuitive to establish

Answer 3

1. A must be related to B
2. A must come before B in time
3. No other variable C can account for the relationship between a and B

No. 3 is the hardest to meet (intuition often skips it)

Question 4

How is controlling for third variables used to argue for causality and what are the limits of this methods?

Answer 4

Impossible to control for all third variables, however, the case for A causing B increases with the more third variables you control for

Question 5

Why is the casual slip so easy to make and how exactly do the media often fail at it?

Answer 5

Media often assumes correlation = causation

We have story making minds, make it very easy to make a casual slip

Question 6

What are the three possible explanations for generational effects?

Answer 6

1. How they were raised (cohort effect)
2. Things going on today that treat younger and older people different (age bias effect)
3. Differences in the amount and quality of life experiences (developmental effect)

Question 7

Know what these are and what problems they create in experiments:

Self selection
Biased dropout rates
Confounds

Answer 7

Self selection:
Study may attract certain types of people, who would more likely put themselves into the experimenter group

Biased dropout rates:
One group has higher dropouts compared to the other. Potentially makes it so the characteristics are different between groups, impacting results

Confounds:
Third variables not accounted for, not controlling for them may lead to significant findings however, A may not cause B

Question 8

Know how identifying a mechanism and specifying the context can strengthen knowledge about casual effects

Answer 8

A casual mechanism: process creating connection between two variables. Many scientist argue that no casual explanation is adequate until a mechanism is identified

Context: no cause has its effect apart from some larger context involving other variables. When, got whom, and in what conditions does this effect occur?

Question 9

How do pre-tests and matching (both individuals and aggregate) help ensure accurate measurement and manipulation of a variable?

Answer 9

Pre-tests: measure the DV before the experiment intervention

Matching: another procedure used to equate experimental and comparison groups, but is a poor substitute for randomisation

Individual matching: Individual cases in the treatment group are matched with similar individuals in the comparison group. Can create comparisons similar to experimental groups

Aggregate matching: in most cases when random assignment is not possible the second method, matching, makes more sense. Identifying a comparison group that matches the treatment group in the aggregated rather than trying to matching individual cases.

Question 10

In general what is a quasi-experiment and when is it used?

Answer 10

Comparison group is predetermined to be comparable to the treatment group in critical ways. It is quasi because they are not randomly assigned to either group.

Used when experiments may be too costly, take too long, unethical, or may be too late to do so.

Question 11

What are the 10 threats to validity?

Answer 11

Selection bias

Mortality (groups become different overtime)

Instrument Decay (measurement instruments wear out overtime)

Testing (Pre-tests can influence post-test scores)

Maturation (Changes in experiment may occur due to other reasons, e.g. subjects may age, gain experience, etc)

Regression (Subjects chosen for low or high scores may suddenly regress or move towards the average)

History (History or external events outside the experiment may influence subjects)

Contamination (experimental and control group may impact each other)

Experimenter expectation (change in experimental subjects may be due to the experimenters positive expectations)

Placebo effect (subject think behaviour should improve, so it will)

Hawthorne effect (experimental subjects results may change due to feeling special)