What are the classical assumptions of linear regression?
mediumAnswer
- (1) Linearity: the conditional mean is linear in the parameters.
- (2) Independence of errors.
- (3) Homoscedasticity: constant error variance across x.
- (4) Normality of errors (needed for exact inference on coefficients, not for point estimates).
- (5) No perfect multicollinearity between features.
- Violations don't ruin prediction but do ruin confidence intervals, p-values, and standard errors.
Check yourself — multiple choice
- Linearity, independence, homoscedasticity, normality of errors, no perfect multicollinearity
- Only linearity and normality
- The features must all be Gaussian
- The target must be strictly positive
The five classical Gauss-Markov / Gaussian-linear-model assumptions.
#linear-regression#assumptions
Practise Supervised Learning
215 interview questions in this topic.
Related questions
- What is VIF and when do you worry about multicollinearity?
- How do you detect and fix heteroscedasticity?
- Why does OLS have a closed-form solution but logistic regression doesn't?
- How do you interpret a coefficient in a multiple linear regression?
- How do you handle strong multicollinearity in a linear model?
- Adding a feature raised R^2. Does that mean the model improved?