4. Multiple linear regression Flashcards Preview

Stats 4 > 4. Multiple linear regression > Flashcards

Flashcards in 4. Multiple linear regression Deck (14):
1

What are the 2 goals of a MLR?

get b values
reduce error

2

What are the 3 types of correlation?

Partial
Semi-Partial
Multicollinearity

3

What are the 2 reasons that multicollinearity is a problem?

It reduces the size of R
It obscures the importance of individual predictors

4

What type of multiple linear regression could you use if multicollinearity is present but you would like to keep the predictors in the model because they explain the underlying construct?

Stepwise

5

What type of MLR has all IVs entered into analysis at once?

Simultaneous

6

How does a simultaneous MLR work?

It assesses each variable as if it had been entered into the regression after all the others

7

What should you generally enter in first in a simultaneous MLR?

The highest correlated (biggest r)

8

How are variables entered into a Hierarchical MLR?

Based on some external criterion of our own, like a theoretical model

9

What are the different types of MLR?

Simultaneous
Hierarchical
Step-wise (forward, backward)

10

In a Step-wise MLR how are variables entered?

Only those variables that contribute statistically to the prediction of the DV

11

When should a step-wise MLR not be used?

Sample less than 40

12

What does a backwards Step-wise MLR do?

Enters everything and then removes those that aren't sig. contributing

13

What does a forward Step-wise MLR do?

Enters IVs with strongest r first and others afterwards only if they can explain sig. further variance

14

Why is backwards step-wise MLR better than forwards?

Forwards = suppressor effect (greater possibility of making Type II error)