You are on page 1of 9

Study Unit 6 : Regression

Multiple Linear Regression


① Exploratory Analysis ( Linear Relationships )
② Correlations
③ Preliminary Regression Model
④ Dummy Variable Regression
⑤ Assumptions ( Normality and Homoscedasticity )
⑥ Multicollinearity
⑦ Find Influential Values
⑧ Model Selection Techniques

Exploratory Analysis

Make use of Boxplots or Scatterplots

Correlations

Man Puts data in a neat table.


Do not have to use it
Preliminary Regression Models

j
① Fit the Model with lm( ) L
② Test Hypothesis : Check if predictor variables are
vs Ha not all zero
significant
③ If Do Not Reject Ho: Stop Analysis
If Reject Ho: Test B’s individually
Fit the Model
Example: ( Do Not Reject Ho )

as

Hypothesis

are
Example: ( Reject Ho )

Fit the Model AND Hypothesis

ARM

Continue Analysis

A
Dummy Variable Regression ( No Interaction )
• Regression for a Qualitative Predictor

Example

A Qualitative ① Create Dummy Variables


② Fit the Model
③ Summary to define the model
④ Graphical Representation

Create Dummy Variables

Fit the Model

OR

O Summary
Graphical Representation

go

my
Dummy Variable Regression ( Interaction )

Example

① Create Dummy Variables


② Fit the Model with interactions
③ Summary to define the model
④ Graphical Representation

Create Dummy Variables

Now the data is


Quantitative

0Fit the Model pm


means interact

OR

Summary

Graphical Representation
Assumptions: Normality and Homoskedasticity
① Extract Residuals
② Extract Fitted Values
③ Graphical Tests
④ Formal Tests

Extract Residuals Extract Fitted Values

Graphical Tests
Formal Tests

I
Multicollinearity ① Basic Correlation Analysis
② VIF (Variance Inflation Factor)
③ Collinearity Diagnostics
The “olsrr” package is needed for the analysis.
Example :

I
Basic Correlation

VIF

Collineary Diagnostics
Influential Values

Model Selection
① Stepwise Selection
② All possible subsets
③ Dummy Variables

Stepwise Selection

All possible subsets

Dummy Variables

You might also like