Multivariate Correlational Research: Longitudinal Designs, Control Variables, and Mediation

These flashcards cover key concepts associated with multivariate correlational research, including designs, causal inferences, mediators, and more.

What is multivariate correlational research?

It's any design that examines more than two measured variables to strengthen causal inferences when experiments aren't possible.

What are the three main approaches used in multivariate correlational research?

(1) Longitudinal designs, (2) Multiple regression, and (3) Pattern and parsimony reasoning.

How does multivariate research strengthen causal inference?

By examining temporal order, ruling out third variables, and synthesizing evidence across studies.

What are the three causal criteria in correlational research?

Covariance, Temporal precedence, and Internal validity.

Define covariance in correlational research.

When two variables are correlated or change together (e.g., overpraise ↔ narcissism).

What does temporal precedence establish?

The cause occurs before the effect in time.

How does internal validity apply to correlational studies?

It involves ruling out alternative explanations (e.g., gender, parental narcissism).

What is a longitudinal design?

A design that measures the same variables in the same participants across multiple time points to clarify directionality.

What is the main strength of longitudinal designs?

They establish temporal precedence and partial internal validity.

What are cross-sectional correlations?

Correlations between variables measured at the same time point.

What are autocorrelations?

Correlations showing stability of the same variable over time (e.g., T1 narcissism ↔ T2 narcissism).

What are cross-lag correlations used for?

Testing whether earlier X predicts later Y and vice versa, providing evidence for directionality.

What does it mean if X → Y is significant but Y → X is not?

X likely causes Y.

What are the weaknesses of longitudinal designs?

Cannot control all third variables; require long-term participation.

What is the purpose of multiple regression?

To test whether a relationship between two variables remains after controlling for other variables.

What is the criterion variable?

The dependent or outcome variable being predicted.

What does the beta (β) coefficient indicate?

The direction and strength of a predictor’s effect after controlling for other predictors.

How does β differ from r (correlation)?

β shows a variable’s unique contribution, while r shows a simple bivariate relationship.

What does it mean if a β’s confidence interval excludes 0?

The predictor has a significant effect.

In the sexual TV content and teen pregnancy study, what did regression reveal?

The relationship remained significant after controlling for factors like SES and age, supporting internal validity.

In the family meals and academic success study, what was discovered after controlling for parental involvement?

The relationship disappeared — parental involvement was the true predictor.

Why can regression not establish causation?

It cannot determine temporal order and only controls for measured variables.

What is pattern and parsimony reasoning?

Synthesizing evidence from multiple studies to find the simplest, most consistent causal explanation.

What does 'parsimony' mean in research?

Favoring the simplest explanation that fits all the data.

How did smoking research illustrate parsimony?

Across studies, smoking consistently predicted cancer; alternative explanations failed, supporting the simplest explanation: smoking causes cancer.

How does media violence research demonstrate pattern and parsimony?

Multiple study types converge to show violent media predict aggression, supporting a causal link.

What does a mediator explain?

The mechanism or process through which one variable affects another (the 'why').

Give an example of mediation.

Conscientiousness → Medical compliance → Health.

What is the ideal way to test mediation?

Measure variables at different time points (X at T1, M at T2, Y at T3).

What statistical methods are used to test mediation?

Multiple regression or Structural Equation Modeling (SEM), often using bootstrapping for indirect effects.

How does a mediator differ from a third variable?

A mediator explains why X affects Y; a third variable creates a spurious link.

How does a moderator differ from a mediator?

A moderator explains when or for whom a relationship occurs, not why.

Give an example of moderation.

The conscientiousness → health link is stronger in older adults.

What is construct validity?

The accuracy and reliability of the measurement (e.g., valid narcissism scales).