Can adjusted R squared be neagtive? If yes, how?
Adjusted r^2 is negative when two conditions are met.
- when R^2 is negative: when your model is Significantly worse than the Base model
if : m = model mean square error and b = base model mean square error
then b >> m
- there is a VERY large difference between the number of observations and the number of features.
number of features are as low as possible.
number of observations are as high as possible.
now if you would pickup the formula of Adj R^2: you will understand what I mean to say.