In the answer to the question: Chi squared goodness of fit test, user glen_b comments that:
The circumstances where it really makes much sense to formally test goodness of fit are surprisingly few.
Perhaps the reason for this that in a lot circumstances, there are better options. However, the one-sample multinomial model appears theoretically appealing, so I wanted to understand:
When does it make sense to practically use the goodness of fit test?
For example a possible use case could be to test (lack of) fit in a logistic model when the cell counts are not too low. However, I have doubts about this. In truth, the few examples like this that I came up with have possibly better alternatives, hence my question.
edit: I'd like to add that if this question does not have a good answer then one wonders why this topic is routinely taught at the undergrad level - even to people who take only one statistics course in college.
