9: Moderation, Regression, Main Effects And Interactions. Flashcards
What type of variables can regression use in moderation?
Categorical and continuous.
What do X axis values have implications for?
Interpretation of regression values: constant and effect terms.
Why are weight values arbitrary?
Changed to the cross-product are matched by changes to the SS for X values.
What does coding matter for?
Interpret
What is contrast coding?
Where the numbers represent meaningful contrasts between groups
What is dummy coding?
The focal group is coded 1 and all others group 0.
What is effect coding?
One group is always -1, each of the others in turn are 1, and the rest are 0.
What does coding not effect?
Means, F , r, and R^2.
Briefly describe a conditional effect.
The impact of Factor A depends on the level of Factor B.
What is implied by an interaction:
Simple effects are different in different parts of the design.
Name 2 things which indicate main effects.
Gaps between lines at the mid-point and the average slope.
Name 2 things which indicate an interaction.
Differences in slopes and differences in gaps (accounting for direction).
What is covered in an omnibus test?
The whole set of conditions, and the overall null hypothesis.
Describe planned comparisons.
Compare 2 individual means at a time.
Describe complex comparisons.
Combine sets of means and then compare them.
What is the name for comparisons which turn everything into a t test?
Single df comparisons.
How is psi computed in planned comparisons?
Sum of products of the contrast weights and the means.
How is SS computed in planned comparisons?
Psi^2 / sum of squared coefficients.
What is the bets estimate of error in planned comparisons?
The MSerror from the omnibus test.
What is calculated to check if comparisons are orthogonal?
The cross-product of the weights.
When are combinations of regression vectors not orthogonal?
When there are more comparisons/ vecors than degrees of freedom for the effect.
How is R^2 max derived from orthogonal sets of contrast vectors?
Summing all r^2 values for the contrast vectors with the DV.
