INTRODUCTION :
Logistic Regression is a regression analysis appropriate to use when dependent variable is dichotomous in nature i.e. binary (0 or 1). It is used when the variable is qualitative or probabilistic in nature.
Example of usage for logistic regression –
Predicting whether a claim for insurance will be raised or not based on factors like age of driver, years of driving experience, gender, vehicle status etc.
Types of Logistic Model :
Probit and logit models are among the most popular models of Logistic Regression:
- Using Logit model,
- Using Probit model,
Transformed Y = NORMSDIST (estimated Y),
( i.e. probability value(z) of estimated Y for a normal distribution with mean = 0 and SD = 1).
Steps involved in Logit Model Maths:
- Find the Estimated Y using linear regression
- Take out the odds ratio of Y=
- Now find the natural log ( Odds Ratio )
- Equate – ln (Odds Ratio) = xβ
- Lastly, solve the equation to find the value of Y as
| Steps involved in Probit Model Maths: |
- Find the Estimated Y using linear regression
- Find out the cumulative probability given the estimated Y using standard normal distribution
- We are finding out the cumulative probability for Estimated Y because -∞ has a probability of 0 and ∞ has a probability of 1.
Example :
Solving Probit Model in Excel:
Maximising the sum using Solver Add – In
Go to Data tab > Click on Solver
