My purpose is to allocate study group randomly by available auxiliary information that are spatial coordinates. I would like to match similiar geographical chracteristics with other data like population of each unit and so on.
I'm not firm on SAS but in ArcGIS you can generate random sampling linked to spatial attributes (sampling within a certain spatial extent). For example in the ArcGIS Toolbox -> Data Management Tools -> Feature Class -> Create Random Points. More detailed description of the tool is availaible here: http://help.arcgis.com/en%20/arcgisdesktop/10.0/help/index.html#//00170000002r000000. Maybe this is something what are you looking for?
I think you could use it to construct the balanced sampler model by yourself modifying the number of samples for each region dependent on the inclusion probabilities (however, for automation a bit scripting effort would be necessary). I haven't tried it yet, but it seems for me to be possible.
If you have some doubts because of scripting, check the ready-for-use tool in ArcGIS Geostatistical analyst, which is aimed to create spatially balanced points. However, Geostatistical analyst is an extension which produce additional costs(!), therefore it is always my last suggestion. But if you have access to it you can use this tool directly.
Here you can find some additional information on this topic: