Logistic Regression Using SAS®: Theory and Application

7.3. Model and Estimation

Now that we’ve seen how to do a discrete choice analysis, let’s look more closely at the model being estimated. Suppose we have i=1, ...,n individuals and each individual is presented with j=1, ...,J_i options. We write the number of possible choices as J_i to indicate that different individuals may have different sets of options. That wasn’t true for the chocolate example, but it’s an important feature of the model that distinguishes it from the multinomial logit model of Chapter 5. Let y_ij = 1 if individual i chooses option j, otherwise 0, and let x_ij be a vector of explanatory variables describing option j for person i. This set of explanatory variables may include dummy variables for the various options and interactions ...