In the practical example, why is the nonlinearity fixed just by logging the variable Y (price)? Would it improve if the other variables / features were also logged?
The opposite mathematical concept of a logarithm is an exponent.
Therefore, when you are observing a relationship which is exponential, you can apply a log transformation to reach a linear relationship.
Alternatively, if you are observing a log transformation, you can apply an exponential transformation to reach a linear relationship.
Note that a linear regression requires a linear relationship and that is why we are trying to achieve it!
The 365 Team
thank you, however I’m still confused on how to identify the type of relationship existing between variables. My question is, when do you know to which variables should the log transformation be applied. Thank you in davance