Hello,
This is more of a general statistics questions than a specific geoprocessing question but reading through the help file of the multivariate clustering tool and searching online I have not found an answer, hence trying my luck here.
I am wondering whether there is a minimum number of features/sample size for multivariate clustering to reveal a sensible result?
Let's consider the example of clustering administrative areas of a country based on their monthly population, i.e. 12 analysis fields (one for every month), into 3 clusters. At level 2 division, there are 200 administrative units and this feels like a sufficient high number for clustering. However, at level 1 division, I might only have 10 administrative units. The tool still runs but will it be a meaningful result with such few units?
Is there a relationship between the minimum number of features required and the number of analysis fields used?
Many thanks