This work presents the sampling procedure used to select the sample got for a "Human Reproduction Study in the District of São Paulo" (Brazil), done by the Department of Applied Statistics of "Faculdade de Higiene e Saúde Pública da Universidade de São Paulo". The procedure tried to solve the situation which resulted from the limitation in cost, time and lack of a frame that could be used in order to get a probability sample in the fixed term of time and with the fixed cost. It consisted in a two stage sampling with dwelling-units as primary units and women as secondary units. At the first stage, it was used stratified sampling in which sub-districts were taken as strata. In order to select primary units, there was a selection of points ("starting points") on the maps of subdistricts by a procedure that was similar to that one called "square grid" but differed from this in several aspects. There were fixed rules to establish a correspondence between each selected "starting point" and a set of three dwelling units where at least one woman of the target population lived. In the selected dwelling units where more than one woman of target population lived, there was a sub-sampling in order to select one of them. In this selection each woman living in the dwelling unit had equal probability of selection. Several "no-answer" cases and correspondent instructions to be followed by the interviewers are presented too.
