Ch8: Bivariate Correlational Research

Association claim

Describes a relationship between variables using verbs like "linked to." Variables are usually measured, not manipulated, so studies are correlational.

Bivariate correlation (bivariate association)

An association involving exactly two variables. It can be positive, negative, or zero.

Bar graph

Used to display a correlation when one variable is categorical. The difference between group means shows the association.

Mean

The average. In a bar graph, the average score for each group.

Mean difference vs. association

Mean difference is how high one group scored compared to another. Association is whether people who score higher on one variable also score higher on the other within a group. Groups can differ in averages while the association inside each stays the same (ex., Dutch people are taller than Americans, but men are taller than women in both).

Effect size

The strength of the relationship between variables (ex., how close r is to 1.0 or -1.0). Larger effects matter more, but most psychological effects are small.

Small effect sizes

Can't be dismissed automatically, since tiny effects can compound when repeated over many occasions or spread across many people.

Statistically significant

A correlation unlikely to come from a population where the true association is zero. In practice, the 95% CI does not include zero (p < .05). If the CI contains zero, it's "not significant," "n.s.," or p > .05.

Replication

Conducting a study again to get multiple estimates of the same association.

Outlier

An extreme score that stands out from the rest. It can make r look stronger or weaker than it really is, especially in small samples. Best found with a scatterplot.

Restriction of range

Not having a full range of scores on one variable, which makes the correlation look smaller than it is (ex., a college admitting only students with SAT scores of 1,200 or higher). Fix by recruiting from both ends or using statistical correction.

Curvilinear association

A relationship that isn't a straight line, such as positive up to a point and then negative (ex., fear and enjoyment in haunted houses). r can be near zero even though a relationship exists, so check the scatterplot too.

Directionality problem (reverse causation)

The temporal precedence issue, where it's unclear which variable came first (ex., happiness might lead to deep talk, not the reverse).

Third-variable problem

An internal validity issue where a lurking variable could explain the association (ex., stress affecting both deep talk and well-being).

Spurious association

A bivariate correlation that exists only because of a third variable (ex., height and hair length correlate only because of gender). Check whether the correlation holds within subgroups.

Criteria for causal claims

Covariance, temporal precedence, and internal validity. Don't infer causation from association alone.

External validity of associations

Depends on how participants were selected, not sample size. A study without a random sample shouldn't be dismissed automatically.

Moderator

A variable that changes the relationship between two other variables depending on its level (ex., residential mobility moderating team success and attendance). If an association is moderated, it doesn't generalize across situations. If not, it does.