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
Covariance:
There must be a correlation between the two variables.
Temporal Precedence:
The cause variable must precede the effect variable in time.
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.