Bivariate Correlational Research

These flashcards cover important concepts from Bivariate Correlational Research, focusing on sampling methods, statistical significance, types of graphs, and the nuances of correlation and causation.

Why does removing certain participants (e.g., 'cautious people') hurt generalizability?

It introduces systematic exclusion, biasing the sample and reducing external validity.

What is an association claim?

A statement that two measured variables are related (knowing one helps predict the other).

What is a bivariate correlation?

The statistical association between exactly two measured variables in the same sample.

What types of directions can correlations have?

Positive, negative, or zero (no relationship).

What is the minimum requirement to test an association claim?

At least two measured variables (no manipulations).

Which validities are emphasized for association claims?

Construct, external, and statistical validity (not internal validity for causation).

Give a plain-language example of a bivariate correlation.

MTL thickness decreases as time spent sitting increases.

When should you use a scatterplot?

When both variables are quantitative.

What does each dot represent in a scatterplot?

One participant’s scores on the two variables.

What features of a relationship can a scatterplot reveal?

Form (linear/nonlinear), direction (positive/negative), and strength (tight/loose clustering).

When should you use a bar graph instead of a scatterplot?

When one variable is categorical and the other continuous.

What are common benchmarks for |r|?

~.10 = small, ~.30 = medium, ~.50 = large.

Why does effect size matter?

Larger effects allow better prediction and are usually more practically meaningful.

Which statistic tests association for two continuous variables?

The correlation coefficient r.

What should you check for construct validity?

Clear operationalization, reliability (e.g., test–retest, interrater), and validity (does it measure the intended construct?).

In the example 'mother’s employment (categorical) and child’s achievement (continuous),' which reliabilities matter most?

Test–retest may matter for employment status; interrater reliability matters for achievement scoring.

What is a confidence interval (CI) around a correlation?

A range estimating the precision of the observed correlation in the population.

What does it mean if a 95% CI for r excludes zero?

The association is statistically significant.

What two factors chiefly influence statistical significance?

Effect size and sample size.

How can outliers affect a correlation?

They can inflate or deflate r, especially in small samples.

What is restriction of range?

Limited variability (e.g., only high scorers) that deflates observed correlations.

Why check for linearity?

r captures only linear relationships; curvilinear patterns can be missed.

Why is replication important?

It confirms reliability of associations across samples and contexts.

Why can’t we infer causation from a single bivariate correlation?

We don’t know which came first (temporal precedence), and third variables may explain the association; correlation satisfies covariance only.

What is a third variable?

An outside factor associated with the predictor that actually causes changes in the outcome, creating a spurious link.

Give an example of a third-variable explanation.

Parental alcoholism explains both early alcohol exposure and teen drinking, undermining the causal claim.

How does gender act as a confound in height–hair-length correlations?

Gender produces the apparent negative correlation; when analyzed within gender, the correlation disappears.

What is a moderator?

A variable that changes the strength or direction of the relationship between two other variables.

How do moderators differ from third variables?

Moderators tell when/for whom a relationship holds; third variables explain why an apparent relationship exists (spurious cause).

Provide a moderator example with stress and depression.

Social support moderates the link: with low support, stress predicts depression more strongly; with high support, the effect weakens.

Provide a moderator example with texting and grades.

Gender moderates the effect: compulsive texting predicts steeper grade declines for girls than for boys.

What chiefly determines external validity in correlational studies?

The sampling method (e.g., random sampling), not sample size.

How can researchers bolster external validity beyond sampling?

Replicate findings across diverse samples, settings, and times.

How should conscientiousness and health checkups be graphed?

With a scatterplot (both variables are continuous).

How should depression status and chocolate consumption be graphed?

With a bar graph (categorical × continuous).

What does a strong correlation look like on a scatterplot?

Points cluster tightly around the regression line, enabling more accurate prediction.

What designs address correlation’s causal limits?

Longitudinal designs, multiple regression (control for third variables), and experiments (random assignment).

Why might a small effect size still matter?

In high-impact or large-scale contexts (e.g., public health, education), small effects can have meaningful population-level consequences.

Define correlation coefficient (r).

A statistic indicating the direction (±) and strength (0–1) of a linear association.

Define spurious association.

An apparent relationship that vanishes when a third variable is controlled.