Multivariate Correlational Designs

• Focus on understanding relationships between multiple variables.

Internal Validity

• A correlational study lacks internal validity.

• Cannot make causal claims based on correlation alone.

• Remember: Correlation is NOT causation.

Establishing Causality

Three Criteria for Causality

1. Covariance:

• There must be a correlation between the two variables.

1. Temporal Precedence:

• The cause variable must precede the effect variable in time.

1. Internal Validity:

• The study design should eliminate plausible alternative explanations (third variables).

Reasons Causal Claims Can't be Made from Correlational Studies

• Directionality Problem:

• Uncertainty about which variable came first; often measured simultaneously.

• Third Variable Problem:

• The presence of a confounding variable that provides an alternative explanation for the observed correlation.

Detailed Explanation of Causality Criteria

1. Covariance

• Establish that the cause and effect co-occur (e.g., A and B are correlated).

2. Temporal Precedence

• Analyze which variable comes first:

• Did A cause B or vice versa?

3. Internal Validity

• Assess if any third variables might influence the relationship between A and B.

Bivariate Correlational Design

• Involves two continuous variables.

• Example: Time spent watching violent shows vs. level of aggression.

Longitudinal Designs

• A multivariate approach measuring the same variables over time to help establish temporal precedence.

• Types of correlations in longitudinal studies:

• Cross-sectional: Correlate two variables at the same time.

• Autocorrelation: Correlation of the same variable over time.

• Cross-lagged: Correlation of one variable at an earlier time with another variable at a later time point.

Ruling Out Third Variables

1. Statistical Control

• Holding a construct constant in analysis to measure the unique effect of a variable.

• Use multiple regression to assess unique contributions of predictors while controlling for third variables.

2. Experimental Control

• Ensuring that constructs are consistent across participants in experiments.

Interpreting Multiple Regression Results

• Significance of Relationship:

• The ongoing link between the predictor and outcome remains significant when controlling for empathy.

• Example implications include predicting increased aggression in individuals who watch more violent TV, even when considering their empathy levels.

• Conversely, if the relationship disappears when accounting for empathy, then empathy may explain the variance in aggression rather than violence in media.

Practice Questions

• Engage actively by answering practice questions via Wooclap.

• Use event code MRSFIA to participate.