REGRESI LOGISTIK UNIVARIAT DENGAN DATA RESPON TIDAK SEIMBANG

Abstract

In binary logistic regression  problem it is common for the two classes to be imbalanced. One case is very rare compared to the other. This paper focus on determining Maximum Likelihood Estimator  on univariate logistic regression for  infinitely unbalanced data response. By finding  the differential of  log-likelihood equation that is resulted, its maximum depend on points    given   through their average  .  Thus we can substitute each  given    with    or substituting all    given     with one   .  If we use    normal distribution for   given     then the slope   can be determined by simple formula   where   and    each an average and variance  for which  .

Keywords : Logistic Regression,  MLE,  Infinitely Unbalanced Data Response