Lecture26 (Data2Decision) Correcting for Heteroscedasticity

[@guroo_ghantaal]

Hi Chris.. Could you please briefly comment on making sense of coefficients after applying box-cox transformations? Thanks for fantastic videos.

[@ChrisMack]

If you need the coefficients to make sense, then a Box-Cox transformation may not be the right way to go. Sometimes, the Box-Cox leads to a simple transformation (a square root, for example), where you can write down the new equation and try to make sense of the coefficients.

[@sichaojiang2284]

Traveling Potato i

[@FadilAidid]

my question is how can we know that Var(ei) = kXi (how do we know this equation comes from?), in my book says exactly the same and it says also there is possibility that Var(ei) = k(Xi)^2.. i am very curious how can we determine that equation?

[@ChrisMack]

The data must show you the way. There are two possibilities. First option: your understanding/theory of the measurements convinces you that the variance of the residuals must follow a specific trend with x. Second, you could plot var(ei) versus xi and see if it displays a trend that makes sense to you. This is easier said than done unless you have a lot of data.