collectStability: Process stability selected networks

In amcdavid/HurdleNormal: 'Estimation and sampling for multi-dimensional Hurdle models on a Normal density with applications to single-cell co-expression'

Process stability selected networks

Description

Stability selection attempts to estimate the probability that an edge would have been included when we sample from the population. Following Shah and Samsworth, we process the stability selection to give stability coefficients for each edge. From Shah and Samsworth: p = #edges = m^2; q = max(iknot) and theta = q/p. We bound the set of population transient edges, which have probabilities of selection below theta (the same theta as above for convenience) when sampling from the population. We call an edge empirically transient when the stability coefficient is below tau. Then for a sparsity theta < .1 and stability coefficient tau > .7, the ratio of empirical transient edges to population edges is less than 1 (Table 2 of Shah and Samworth)

Usage

collectStability(stabout, theta = 0.1, stabilityCheckpointDir = NULL)

Arguments

stabout	output of stability
theta	sparsity of the solution in terms of the number of parameters. default .1.
stabilityCheckpointDir	path to a directory to write checkpoint files

Value

A list with components. You probably want component 3, giving the stability coefficients for each edge.

amcdavid/HurdleNormal documentation built on May 14, 2022, 11:12 p.m.