Regression Analysis

Question 1

Two ways of making a prediction

Answer 1

Extrapolation: Prediction based on the past consistent pattern

Predictive Modelling: Predictions based on the relationships with variables

Question 2

Goodness of a prediction

Answer 2

Differences between predicted and observed values

Question 3

Bivariate Regression Analysis

Answer 3

y = a + bx

1) Basis of “least square criterion”
2) R square: how well the straight line model fit the observed points

3) Testing the regression model:
- Ho of regression model (F-Test): No linear relationship between DV and IV’s
- Ho of each IV: No linear relationship between DV and each IV

Question 4

Multiple regression analysis

More than 1 IV

Answer 4

R square
Test for the overall regression model (Ho for overall model)
t-test for each coefficient (Ho for each IV)
Multicollinearity
Special uses for Multiple regression analysis
Stepwise regression
Dummy variables

Question 5

Multicollinearity

Answer 5

If VIF (Variance Inflation Factor) > 10 or tolerance is close to 0, multicollinearity is suspected.

How to fix the problem
Examining the correlation matrix, drop one
Taking logarithm

Question 6

Special uses of Multiple regression analysis

Answer 6

Screening Device 
Standardized betas (used for ranking IVs in terms of their importance)

Question 7

Stepwise regression

Answer 7

Successice entry of IV’s based on p-Values