10 Interactions in GLM: categorical and continuous variables Flashcards

1
Q

What question is addressed by interaction effects?

A

Does the relationship between the IV and DV differ depending on the level of another IV.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

If the relationship between an IV and a DV differs depending on another variable, that relationship is?

A

Heterogeneous. There is heterogeneity of covariance.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What two transformations are needed when calculating an interaction effect using continuous variables?

A
  1. Mean-centre the variable.

2. Calculate the cross-products.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What are two advantages of mean-centring?

A
  1. Constant is immediately interpretable, as 0 is mean of that score. E.g. if using intelligence as a predictor of Y, the constant is the score on Y for someone of average intelligence.
  2. Avoiding multicollinearity. E.g. If multiplying positive variables to create an interaction variable, high numbers will go higher and low will stay low –leading to correlation with first-order variables.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What’s a simple way of getting simple bivariate regressions from a categorical variable? And what are the limitations?

A

Split the file to see what is going on with each group.

Limitations:

1) can see that groups are different, but cannot test whether this difference is significant
2) Not using df from the whole sample.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What does the squared semi-partial (or part) correlation tell us?

A

The unique contribution
of that variable in model –over and above everything else. If we put that variable in a hierarchical model, this would be the same as R^2 change.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

In multiple regression, an interaction beta coefficient represents what?

A

The difference between the regression slopes of the two original variables.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

What are two ways of doing a simple slopes analysis (getting a Beta coefficient) with a two-level categorical variable?

A
  1. Dummy code the variable twice, making each level of interest zero (i.e. the reference group).
  2. Compute two new variables, splitting up the OTHER predictor variable (IV2). In the first, set all values on IV2 to 0 for level 1 of IV1, and identical to IV2 for level 2.
    In the second, copy all values on IV2 for level 1 of IV1 and set them to zero for level 2 of IV1.
    Enter these variables into a regression with IV1 (but not IV2).
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What is the test of a simple slope testing?

A

Whether each group’s slope is different from zero.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

When you centre variables, the _______ _______ don’t change, but the _______ does.

A

When you centre variables, the regression coefficient doesn’t change, but the constant does.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

In regression, what does a significant interaction mean?

A

That the relationship between Y and X1 changes with different levels of X2.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

How can you split a continuous variable into low, middle and high groups?

A

Sub in + or - 1SD to the regression equation as the value of that variable (and of the interaction variable if there is one).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What does a non-significant interaction mean, given that main effects are significant, in a 2x2 design?

A

That there is no difference between the differences.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly