Regression model assumptions

Hi all,
I think it would be interesting to show an end-to-end example with a dataset that comes from real data.
For example, there are certain assumptions (Normality, Zero Mean, homoscedasticity) before you can apply a regression model. I have noticed that many people blindly apply this model (including myself, a few years ago) but it is mathematically incorrect. Code snippets to perform assumption checking tests would also be useful for students. What do you think about that?

1 Like

Hey @Degeye,
Welcome to the community. There are 2 points of view. For a learner who is just beginning to learn Machine Learning, starting off with something easy without these assumption checks, might be more preferable, in order to make sure the learner grasps the bare essentials. But for a person who is already familiar with the basics of Machine learning and the various common algorithms, it would be better that if these assumption checks are included, since such a learner would be looking for something extra.

However, in both the cases, a note regarding these assumptions should definitely be included, and perhaps an optional lab for the learners who are looking for extra.

Let me suggest this to the course team, and let’s see what they decide. Thanks for your input.



Hi @Elemento,
I totally agree with your point of view and I thank you for your proposal to add an optional part because it is a step that will condition the choice of a model and it is important to see what are the triggers for the acceptance or not of a model choice.

I know that Professor Andrew Ng is very attentive to the fact that AI does not remain at a purely pedagogical stage but also applicable in the real world in different fields of application (

We can see that there are people who come from the world of Statistics and others from the world of Machine Learning. These two worlds are not mutually exclusive, but it is common for each to stay in their original domain. I think this is an additional argument to allow people from the Machine Learning world to be aware of the link that exists. A student taking the current course could very well be led to use a regression model without knowing the initial conditions.

Have a nice day!

1 Like