In order to estimate the score function parameters over Ω∗,
we use the data on hand of the two subpopulations. The use of
customer subpopulation Ω data aims to moderate the small size
of the subpopulation Ω∗ of non customers, by supposing the
existence of hidden links between the distribution of variables
over Ω and that over Ω∗.
It’s known from [20] as well as [18] that existence of particular
connections between the variables distributions lead to
relations between the parameters of their respective logistic
regression models, consequently our task consists in finding
these links. In this context a preliminary case study was
successfully done in Gaussian multivariate case [19]. It is a
question here of extending the found results in Gaussian case
to logistic case, which leads to simple and parsimonious linking
models between the parameters of logistic classification
rules associated respectively to the two subpopulations Ω and
Ω∗.