I am reading Introduction to Statistical Learning and it is said (as in other websites) that Logistic Regression is unstable compared to Linear Discriminant Analysis in well separated cases. Specifically, on p. 138 in the 7th edition:
When the classes are well-separated, the parameter estimates for thelogistic regression model are surprisingly unstable. Linear discriminant analysis does not suffer from this problem.
Can you please provide a simple example why this is so?