week 3 - introducing interactions into multiple regression models Flashcards
1
Q
what is model comparison
A
- an approach to multiple regression is to create several models each building upon each other
- nested models
- the latest model can differ by +1 predictor
2
Q
what models do you use
A
3
Q
what are the problems here
A
- can lead to p-hacking or HARKing
- state clearly whether you are in confirmatory or exploratory analysis mode
- explain process
- report all models
- be honest and comprehensive
4
Q
what shows simple regression
A
5
Q
what is shown in the presence of two groups
A
6
Q
how do you show the different rates of change across levels
A
7
Q
what shows a interaction
A
8
Q
what are interactions
A
- when one variable moderates the effect of another variable on the outcome variable
- when a coefficient for one predictor changes its size when it comes between the levels/values of another predictor
- the change that one predictor predicts for the outcome variable depends on a specific level of another predictor
9
Q
what is moderation
A
- a process for exploring the differing conditions under which two or more variables may work together
10
Q
what are interaction terms
A
11
Q
what are the equations for multiple regression
A
y = b0 + b1 * X1 + b2 * X2 + b3 * (X1 * X2) + e
12
Q
when are interaction terms used
A
- theory my predict them
- study design may implicitly suggest them
- results may suggest them
13
Q
what is the statistical power of interactions
A
14
Q
how do you build a model with interactions
A
- model type is based around research question/hypotheses/theory
- if this includes an interaction term then build first model
- if not build an independent predictor model first
- build further model that includes independent and interaction terms
