Question

The logit model specifies P(Y=1|

  • X = Λ(Xβ). OLS is inappropriate for two fundamental reasons. Which correctly states BOTH?
A OLS requires normality of errors and a linear functional form; both violated since logit errors follow a logistic distribution and the link is non-linear
B OLS is inconsistent for logit because: (1) binary Y produces heteroskedastic errors [Var(ε|X) = P(X)(1−P(X))], violating Gauss-Markov; AND (2) OLS predictions are not constrained to [0,1], making them theoretically incoherent as probabilities
C OLS cannot estimate logit because the log-likelihood has no closed-form solution
D OLS gives biased estimates because the logistic CDF is asymmetric while OLS assumes symmetric residuals
Practice Next

Hey! Ask a query