PSYC3010 - Course Review Notes

Course Review

Factorial Designs
Factorial Between-Participants ANOVA
Correlation & Standard Multiple Regression
Standard & Hierarchical Multiple Regression
Moderated Multiple Regression
Mediation & Indirect Effects
Within-Participants ANOVA
Mixed ANOVA: Research questions, variance, omnibus tests, follow-up tests, advantages.

Designs with one IV/factor/predictor: Focus on direct relationships; does the IV affect the DV?
Designs with multiple IVs/factors/predictors: Focus on direct relationships + interactions; do IVs interact in affecting the DV?

ANOVA is used to analyze experimental designs with categorical groups/conditions.
Regression is used to analyze correlational designs with measured variables.
Design and analysis are not the same thing.
There are various approaches to analyze data from specific designs.

Compares group means to see if they significantly differ on a DV.
Used with categorical IVs/factors/predictors.
Appropriate for manipulated groups in experiments or measured group memberships.
Represents effects in terms of mean differences.
Tests whether variance in DV is explained by IV(s) by examining differences between means.
Partitions variance into explained variance + error/residual.
Quantifies variance in DV each effect accounts for (\eta^2, \eta_p^2, \omega^2).

Tests linear relationships between variables.
Can be used with continuous IVs/predictors.
Can also be used for categorical IVs/predictors by numerically coding the groups.
Represents effects in terms of linear relationships.
Tests whether variance in criterion is explained by predictors(s) by examining linear relationships.
Partitions variance into variance explained by predictors + error/residual.
Quantifies variance in DV the overall model (R^2, R_{adj}^2) and each predictor (pr^2, sr^2) accounts for.

Experimental research: Examines relationships between manipulated IVs/factors and measured DVs; allows causal inferences.
Correlational research: Examines relationships between measured variables; does not allow causal inferences.

Methodological designs should not be conflated with statistical analyses.
Causality comes from experimental manipulation + random assignment.
Both ANOVA and regression can be used to make causal inferences for manipulated IVs.
Neither ANOVA nor regression provides causal evidence for measured IVs.

ANOVA and regression are based on the same underlying model.
GLM models a numeric outcome variable as a linear combination of predictor variables.
Uses linear equations to describe patterns in data and make predictions.
Example: Structural equations underlying two-way ANOVA and MMR.
X{ijk} = \mu.. + \alphaj + \betak + \alpha\beta{jk} + e{ijk} Y = a + b1X + b2W + b3XW + e