permPower {GraphAT} | R Documentation |
This function simulates graphs from the alternative hypothesis of preferential connection of intracluster nodes. For each graph, it runs a node and edge permutation test. The estimated ``power'' of each test is the proportion of graphs that the test rejects the null hypothesis of no preferential connection of intracluster edges.
permPower(psi=1,clsizes, nedge, nhyper=100, nperms=1000)
psi |
The non-centrality parameter for the noncentral hypergeometric distribution used to simulate the graphs. |
clsizes |
A vector of cluster sizes. |
nedge |
The number of edges in each graph. |
nhyper |
The number of noncentral hypergeometric graphs simulated to estimate "power". |
nperms |
The number of permutations used for each run of the edge and node permutation tests. |
The function first generates nhyper realizations of a noncentral hypergeometric(nedge,n,k,psi) random variable, where n is the number of node pairs and k is the number of intracluster node pairs. For each realization x, a graph with n edges, x of which are intracluster, is generated. The edge and node permutation tests (with nperms permutations each) are performed on each graph. The estimated ``power'' of each test is the proportion of graphs for which the test rejects the null hypothesis of no preferential connection of intracluster nodes (at the 5% level). The 95% confidence intervals for the power levels are also computed.
A list with four components:
power.permedge |
Estimated ``power'' for edge permutation test. |
power.permnode |
Estimated ``power'' for node permutation test. |
CI.permedge |
Vector giving 95% confidence interval for edge permutation test power. |
CI.permnode |
Vector giving 95% confidence interval for node permutation test power. |
Tom LaFramboise tlaframb@hsph.harvard.edu
permEdgesM2M
, permNodesM2M
, makeClustM
permPower(psi=5,clsizes=c(1,2,3,4),nedge=10,nhyper=100,nperms=100)