It is a form of non-probability sampling in which the researcher utilises some knowledge about the population to build some represkntativeness into the sample plan. The population is divided into several subgroups according to the distribution of population in these subgroups. Then, based on the proportions of distribution of population in the subgroups the number of sample units from each subgroup are selected. To illustrate we can take the known distribution in the proportion. You know that women within reproductive age group 15-45 years is 17% of the population. Thus if you are doing an opinion poll on family planning practices your sample should have 17% from the women of 15-45 years.