Selecting a random subset of grid cells

[JoeBanker]

I need to select a random subset of spatially independent grid cells for a species occupancy modelling study.  I defined my sampling units as 100x100m grid cells covering the entire study area. However, these grids are continuous and not independent from one another (see attached). I would like to select a random subset of grids (let's say 50%) which is at least 100m from the closest grid (between each grid cell there must be at least a 100m distance). Any advice on this will be much appreciated, also, I'm using ArcGis 10.6.

[DuncanHornby]

There are two approaches to this problem:

Select random rows from tables

You could use the code here to randomly select rows in your grid layer, assuming they are polygons.

However this does not impose the spatial criteria of at least 100m separation distance, so you would need to adapt the code to test for this scenario and adjust accordingly.

Create random points

An alternative approach is to use the Create Random Points tool and ensure your polygon layer is the constraining featureclass and set the minimum allowable distance. You can then use those points to select the grid squares.

Notes

Something to be aware of is that when imposing a spatial criteria this could mean you never achieve the desired final number selected. 

Also not being a statistician guru I'm not sure which approach minimises bias, may be someone else  can comment?