I am working on a statewide project to create surface water quality grids from point samples. In some instances, we have large sample sizes (n>50,000) which is problematic when creating semiovariograms in ArcGIS' Geostatistical Analyst. See my sample locations for groundwater pH. I'm looking for some literature on how to sub-sample my data prior to interpolation so I can look at the semiovariogram, trends and geostatistics.
From what I understand of your desire, there are many ways you can go about this. One example would be to use the Subset Features tool. You can create a random subset based on your desired percentage of the original. If you want to be more python-y, the top answer here has a good example. You can also check you the add-in tool here: http://www.arcgis.com/home/item.html?id=ecbe1fc44f35465f9dea42ef9b63e785
