You are on page 1of 8

Quantitative Techniques in Management I

Session 25
Reading: SfM Ch. 15

Logistic Regression
Used in multiple linear programming to identify the

probability of an event occurring, used when the


dependent variable y is a categorical variable.
Logistic regression:
The goodness of fit is measured by calculated the

Deviance statistic, and comparing it against the chi-square


chart, df=n-k-1.

Nonlinear relationships: Quadratic


regression

The predicted value


Testing for significance: (no relationship between and ),
Use F-test, F=MSR/MSE, df=(2,n-2-1)
Test for quadratic effect:
T-test,

Transforming Regression Models


Models are transformed to overcome violations in

assumptions of regression models, or to convert a


nonlinear model to linear form.
Sq. Root. Transformation:
Log transformations:
Original Multiplicative model:
Transformed:
Original exponential model:
Transformed:

Collinearity
Collinearity

occurs when two or more independent variables are

highly correlated with each other. It becomes difficult to separate


the effect of collinear variables on the dependent variable.
Let be the coefficient of multiple determination where is
treated as the dependent variable, and all other x are treated as
independent.
The Variance Inflationary Factor
When is considered excessively collinear, and least-squared
regression is inadvisable. Results can only be predicted using
independent variable values that are within previously observed
ranges.
The issue may be alleviated by eliminating the variable with
highest

Model building
There is one dependent variable and many possible

independent variables
Parsimony dictates that we use the regression model with
fewest variables.
First, evaluate the VIFs of all independent variables, and
eliminate those with high collinearity
Then use stepwise regression. First, use partial F-test or ttest to see which model with one independent variable is
the best fit.
Then, choose a second independent variable (all tests
normally at 5% sig.) which shows a positive contribution to
the model, and has the largest contribution.

Best-subsets approach to model


building
Evaluate every possible combination of independent

variables to choose most appropriate model.


is the coefficient of multiple determination for the
regression model containing all T parameters (dependent
and independent)
is the coefficient of determination for regression model
with k independent variables ()
Calculate
Identify those models whose , these are the acceptable
models, then use some other method like Parsimony to
decide the best one.

Look at the data in Session 24.xlsx. Subset-regression

shows that the best option is to use all three dependent


variables. In Stepwise regression, the first variable to be
added is X3, followed by X2, then X1.
VIF-analysis shows that removing X1 (CO2 emissions)
greatly reduces collinearity.

You might also like