As anticipated, the linear function is not suitable to represent the relationship between the independent variable and the dependent variable, in cases where the response takes dichotomous values. Now, we try to model the probabilities that the response belongs to the reference class. If we use the standard linear model to represent these probabilities, then our system is described by the following equation:
In this case, we have predictions that both classification options may exceed the expected values (-∞ to +∞), as we could see from the figure shown in the previous section. Then, to solve this problem we can apply an exponential ...