If you're interested, in the scientific literature this problem is known as "network backboning". Basically you have all nodes connected to practically all othe

If you're interested, in the scientific literature this problem is known as "network backboning". Basically you have all nodes connected to practically all other nodes with weighted edges, and you want to know which are the edges with statistically significant weights. I wrote on this topic [1]. My method [2] basically uses simple counts on edge weights, and then estimates the expected edge weight and its variance using Bayesian priors. It then attaches a t-score or p-value to each edge, and then you can filter out edges with too low t-score. The idea is that weak edges can still be statistically significant if they connect "small" nodes. In any case, the library I wrote includes the implementation of a few other methods, in case they work better for your data type. [1] https://arxiv.org/abs/1701.07336 https://arxiv.org/abs/1701.07336 [2] http://www.michelecoscia.com/?page_id=287 http://www.michelecoscia.com/?page_id=287