Stratification is frequently used in sampling. In human surveys, a typical stratification exploits administrative areas or regions of the frame population. Sometimes, gender and age have also been used. The paper continues to stratify a sample regionally, but in two ways. One sample is now a typical administrative area composition, that is, the regions are municipalities or sub-municipalities. This sample aims at examining well the whole target population, their living conditions and attitudes. But we have another target at the same time. We want to compare living conditions and attitudes between rich and poor areas. For this purpose, we use grid 250 m x 250 m squares. The Finnish taxation data are used for classifying all the grid squares of the target population into the five equally size categories (quintiles) by average income, and then two strata are formed, ones constituting the lowest quintile incomes and the second the highest income, respectively. Finally, the sample is drawn at random from both strata. It is expected that there are in a sample more poor people in the lowest income stratum and consequently more rich people in the highest income stratum. Since we have also in the first sample people who can belong to all types of income grid squares, we have opportunity to analyse well the whole population and its sub-groups. This requires for example to create the single sampling weights for the entire data. The methodology for this weighting will be presented too.