Dummy variables
A dummy equals 1 if a condition holds and 0 otherwise: a recession quarter, a sector, a policy change.
- Intercept dummy: Y = b0 + d0D + b1X. When D = 1 the line shifts up by d0.
- Slope dummy: Y = b0 + b1X + d1(D × X). When D = 1 the slope becomes b1 + d1.
- Both together let one group have its own intercept and slope.
For n categories, use n − 1 dummies. The omitted category is the base, and each dummy coefficient is the difference from it.
Influential observations
- High leverage: an observation with extreme X values. Leverage hii above 3(k + 1)/n is a common flag.
- Outlier: an extreme Y given its X. A studentized residual above the critical t-value flags it.
- Cook’s distance: combines both. Values above √(k/n) × 2, or above 1, suggest the observation is influential.
Influential points can pull coefficients a long way. Check whether they are errors before acting.
Logistic regression
For a binary dependent variable (default or not, takeover target or not):
ln[p ÷ (1 − p)] = b0 + b1X1 + … + bkXk
p = 1 ÷ [1 + e^−(b0 + b1X1 + … )]
- Coefficients are changes in log-odds. e^b gives the factor by which the odds change for a one-unit increase in X.
- Estimated by maximum likelihood.
- Significance: z-tests on coefficients, a likelihood ratio test for the model, and pseudo-R² to compare models on the same data.
Example: log-odds = −0.6 → odds 0.549 → probability 35.4%.
Exam focus
Set up dummy variables correctly, interpret intercept and slope dummies, identify leverage points and outliers, and convert logistic regression output into probabilities and odds ratios.