1. Forecasting Choices

2. Types of Variable Variable Quantitative Qualitative Continuous Discrete (counting) Ordinal Nominal

3. Nominal or Ordinal Dependent Variable Indicating “choices” of a decision maker, say a consumer. Response categories: –Mutually exclusive –Collectively exhaustive –Finite Number Desired regression outputs –Probability that the d.m. chooses each category –Coefficient of each independent variable

4. Generalized Linear Models (GLM) Regression model for a continuous Y: Y =  0 +  1 X 1 +  2 X 2 + e  e following N(0,  ) GLM Formulation: 1.Model for Y: Y is N( ,  ) 2.Link Function (model for the predictors)  =  0 +  1 X 1 +  2 X 2

5. Estimation of Parameters of GLM Maximum Likelihood Estimation –For normal Y, MLE is the LS estimation Maximize: –Sum of log (likelihood function), L i of each observation

6. MLE for Regression Model Y is N( ,  ) MLE: Maximize

7. GLM for Binary Dependent Variable, Y Model for response: Y is B (n,  ) Model for predictors (Link Function) logit(  0 +  1 X 1 +  2 X 2 +…  K X K = g Probability   exp(g) / (1+exp(g))

8. X : Covariates Independent variables are often referred to as “covariates.” Example: –SPSS binary logistic regression routine –SPSS multinomial logistic regression routine

9. A. Logistic Regression For Ungrouped Data (n i =1) Model of Observation for the i-th observation Y i = 1: Choose category 1with probability  i Y i = 0: Choose category 2with probability 1-  i Log Likelihood Function for the i-th observation

10. MLE Maximize:

11. Setting Up a Worksheet for MLE Define an array for storing parameters of the link function. Enter an initial estimate for each parameter. Then for each observation: Sum the likelihood and invoke the solver to maximize by changing the parameters. Multiply –2 to the maximized value for test of significance of the regression Link Function, g i Parameters of the Likelihood ln(Likelihood) L i

12. Test of Significance Hypotheses: H 0 :  1 =  2 ….   = 0 H 1 : At least one  j = 0 Test statistic: The Distribution Under H 0 :   (DF = K)

13. Standard Errors of Logistic Regression Coefficients (optional) Estimate of Information Matrix, I (K=2)

14. Deviance Residuals and Deviance for Logistic Regression (Optional) Deviance (corresponds to SSE) Deviance Residual

15. B. Logistic Regression for Grouped Data Using WLS The observation for the i-th group: ->

16. WLS for Logistic Regression Regress: on X 1i, …, X Ki with

17. WLS for Unequal Variance Data X Y * * * * * 1 2 Observation 2 is subject to a larger variance than observation 1. So, it makes sense to give a lower weight. In WLS, the weight is proportional to 1/variance.

18. Modeling of Forecasting Choices - GLM 1.Model for Observation of the Dependent Variable. A probability distribution Link Function (Model for Independent Variables) A mathematical function

19. Forecasting Choices # of Choices 2 Binomial Distr. > 2 Multinomial Distr. UnorderedOrdered

20. Multinomial Logit Regression Multinomial Choice (m=3), Ungrouped Data: –Y 1 =1: Choose category 1with probability   –Y 1 =0: Choose category 2 or 3with probability 1-   –Y 2 =1: Choose category 2with probability   –Y 2 =0: Choose category 1 or 3with probability 1-   –Y 3 =1: Choose category 3with probability   –Y 3 =0: Choose category 1 or 2with probability 1-   

21. Log Likelihood Function Log Likelihood Function of the i-th ungrouped observation MLE: Maximize

22. Y 3 and  3 can be omitted Multinomial Choice (m=3), Ungrouped Data: –Y 1 =1: Choose category 1with probability   –Y 1 =0: Choose category 2 or 3with probability 1-   –Y 2 =1: Choose category 2with probability   –Y 2 =0: Choose category 1 or 3with probability 1-   

23. Log Likelihood Function Log Likelihood Function of the i-th (ungrouped) observation MLE: Maximize

24. 1: Formulating “Link” Functions: Unordered Choice Categories Category 3 as the baseline category.

25. From Link Functions to Probabilities

26. Test of Significance Hypotheses: H 0 :  11 =  21 = …  K1 =  12 =  22 = …  K2 = 0 H 1 : At least one  ij = 0 Test statistic The Distribution Under H 0 :   (DF = 2 K)

27. Interpreting Coefficients Not easy, as a change of probability for one category affects probabilities for other (two) categories.

28. 11 22 2: Formulating Link Functions: Ordered Choice Categories Underlying Variable Defining Categories Category 1Category 2Category 3

29. Choices for Probability Distribution of U a. Ordered Probit Model for the i-th DM U i = follows N(  i,  =1) b. Ordered Logit Model for the i-th DM U i follows Logistic Distribution(  i )   i =  1 X 1i +  2 X 2i (no const)

30. a. Ordered Probit Model

31. b. Ordered Logit Model

32. Types of Variable Variable Quantitative Qualitative Continuous Discrete (counting) Ordinal Nominal

33. Poisson Regression for Counting Model of observations for Y Link Function Log Likelihood Function