Let's say that we have two classes, A and B, that we are trying to predict. Let's also suppose that we are trying to predict A or B based on a variable x. Classes A and B might look something like this when plotted against x:
Although we could draw a line modeling this behavior, this is clearly not linear behavior and does not fit the assumptions of linear regression. The shape of the data is more of a step from one class to another class as a function of x. What we really need is some function that goes to and stays at A for low values of x, and goes to and stays at B for higher values of x.
Well, we are in ...