Explanation

This question tests knowledge of common regression problems and violations of classical linear regression assumptions.

Common Regression Problems:

1. Multicollinearity - When independent variables are highly correlated with each other

   • Violates assumption of no perfect multicollinearity
   • Causes unstable coefficient estimates and high standard errors

2. Heteroscedasticity - When error terms have non-constant variance

   • Violates assumption of homoscedasticity
   • Leads to inefficient estimates and incorrect standard errors

3. Autocorrelation - When error terms are correlated across observations

   • Violates assumption of no serial correlation
   • Common in time series data

Stratification is NOT a regression problem:

• Stratification is a sampling technique where the population is divided into subgroups (strata)
• It's used in survey sampling and experimental design
• It's not a violation of regression assumptions

Key Regression Assumptions (CLRM):

1. Linearity in parameters
2. Random sampling
3. No perfect multicollinearity
4. Zero conditional mean
5. Homoscedasticity
6. No autocorrelation
7. Normality of errors (for inference)