"The problem we are analysing arise when we utilise tests ""screening"" to evaluate the conditions (to diagnose the aptitude) of a number of candidates to receive a mortgage from a financial entity, with the problems of lack of information, more specifically, supposing we have two tests with two possible results respectively called (R+) and (R-);the results of both tests are conditionally independent and the possibility of obtaining a false positive results for each test is zero. As the objective of this study we propose to utilise a methodological algorithm based on the statistic technique known as E.M., to determine the probability of concession of the credit and to know which test is more reliable. We present the algorithm E.M. and its relationship to the method of Maximum Likelihood and the parametric estimation, and the use of this algorithm to solve the problems which arise when the data is incomplete or censored. consisting of two phases such as complete data relating to data observed or incomplete, following which we define the function of Likelihood to which we calculate its Expected value phase E and then we maximize the above mention value, phase M. Finally we present a practical case in which we realize all the process of the algorithm E.M., the different expressions obtained being programmed and where apart from some initial values we obtain the different values which give us the probability of concession and so, which is the best test to decide the concession of the mortgages."