Hurdle Model
The hurdle model separates edge presence from conditional positive edge strength. The presence probability is modeled as a logistic function of a global intercept and the latent edge mean. The coupling parameter determines whether absent e…
1 sources - 4 claims
The hurdle model separates edge presence from conditional positive edge strength. The presence probability is modeled as a logistic function of a global intercept and the latent edge mean. The coupling parameter determines whether absent edges are informative about weak latent connectivity. Conditional nonzero weights are modeled with a Gaussian likelihood, with a Student-t alternative available for robustness.