Suppose the underlying true graph $G^T$ is generated from some known random graph model. We are able to obverse a partial graph $G^O$ (Assume the difference between $G^T$ and $G^O$ is that some links are missing in $G^O$).
Based on this $G^O$ and our knowledge about the random graph model, can we estimate the likelihood of links among each pair of nodes in $G^O$? I wonder if there is a systematical way/related research to treat this problem.
I am aware that there are some techniques like link prediction to predict missing links; but it only uses the topology information of $G^O$ without the knowledge of the random graph model.