Linear Regression assumptions not satisfied



Hi All,

What should we do if we are using linear regression for a problem(target variable is continuous) and the linear regression assumptions are not satisfied. Should we use a different algorithm or modify the variables so as to comply with the assumptions.

Poorna R



This is a good question, but it would need more details - which assumption is not being satisfied? To what degree? It would be helpful, it you can include some of the dispersion plots of residuals / variables.




I was asked this question by someone , if the normality and equal variance assumption of the errors is not satisfied, what would we do in such a case.

Poorna R


If these assumptions are not satisfied we go for transformation of indepwndent variables like box cox transformation, tukeys transformation etc


Sorry not independent, we do tramsformation of dependent variable