I have samples of individuals from various areas in South and Central America (12 samples, so 12 locations). Each sample has a different number of individuals with a measurement assigned to them. Each sample also has somewhere between 4-40 individuals. So there are coincident data at each location.
I would like to see if there is spatial autocorrelation (clustering/dispersion) among the samples taking into account this measurement. I am using the Spatial Autocorrelation Moran's 1 tool. Having done something similar to this before, I felt each location could only have one value. In the past (using geostatistical analyst), it would just average the values at each location. So I went ahead and did that myself. Thus I had 12 locations, each with one data point. I ran the spatial autocorrelation tool, and there was nothing significant.
I am doing it again without averaging the samples. Thus each location as between 4-40 data points. After running the tool again, it finds significant clustering. However, if there are 12 locations with a ton of data points, I feel this is due to sampleing bias. Is that right?
Any suggestions on the best way to do this?