In August, 1967, our research group which was composed of eleven members headed by Syoichi Nakamura carried on a research about the actual conditions and opinions of the inhabitants of Airin Area (Kamagasaki) in Osaka. We picked up spots by replicated sampling design and interviewed all the adults, when possible, living there.
We could employ family or person as a sampling unit instead of spot, but in the case of Airin Area research based on the registered records, which are usually used for the frame of the population in this sampling unit, such as resident registration cards or lists of voters filed at the ward offices would produce a considerable divergence to the reality.
Building was another sampling unit we considered. But to make sure the unit, it would be necessary passable preparatory research because we didn't have any sufficient maps to be based on. And that there left some problems in practicing the method, and after all we decided to employ spot as the sampling unit.
We had already considered random sampling of spots in 1959 when we had researched Airin Area for the first time. At that time, we selected the typical spots and tried to define the characteristics of the spots. After that collection of data on each problem followed, but the sampling research of the inhabitants all over the area had not been done.
Sampling of the spots were performed as follows. First, we drew parallel lines horizontally and vertically at intervals of 5 millimeters on the maps (1 : 1500, published by Japan Resident Association 'Nihon Jutaku Kyokai') of seven selected towns, smoothening the circumferences, and got 4982 cells. Then we devided these cells into 50 zones, each of which contained 100 cells. From each zone we picked up 10 sub-samples by random sampling.
We employed this method (1) because estimation of standard errors was easy, and (2) because in the case of cutting the personnel or accidental interruption of the reseach, this research would not lose the significance as a sampling tecnique itself.
We got the result that the number of families was 1.52 times as much as the registered records and the population 1.62 times. But this research resulted static because the time control of the sampling was insufficient for the mobile subject like the inhabitants of Airin Area.