14.3 Other Issues and Problems with Data

More serious problems than heteroscedasticity:

  • Functional form mis-specification
  • Measurement error
  • Missing data: Estimating a standard regression model is not possible with missing values. All statistical software packages ignore missing data. Missing data is a minor problem if it is due to random error. Missing data can be problematic if it is systematically missing, e.g., missing education data for people with lower education
  • Non-random samples
  • Outliers