crossval: Cross-validate network on new data

In ghosoya/ludwig: Simulating and Estimating Undirected Binary Networks Using the Pseudolikelihood Method

Description

This intended use of this function is helping cross-validate an estimated network on new data. To this aim, the values of individual nodes are predicted based on the conditional predictive distributions and for each node Cohen's Kappa, sensitivity, specificity, prevalence and accuracy statistics are provided as well as the frequencies of true positives (TP), true negatives (TN), false positives (FP) and false negatives (FN).

In addition, Cohen's Kappa, sensitivity, specificity and accuracy of the predictions are reported for all nodes or, in other words, the whole network.

Usage

crossval(net, dat, debug = FALSE)

Arguments

net	An object of class estnet
dat	A validation dataset
debug	Debug information

Value

The function returns a list of class crossval, containing...

• kappa Cohen's Kappa for each item.

• sensitivity Sensitivity of the predictions for each item.

• specificity Specificity of the predictions for each item.

• accuracy Accuracy of the predictions for each item.

• prevalence Prevalence of the predictions for each item.

• TN Number of true negative predictions for each item.

• FN Number of false negative predictions for each item.

• FP Number of false positive predictions for each item.

• TP Number of true positive predictions for each item.

• kappa_total Cohen's Kappa for the whole network.

• sensi_total Sensitivity for the whole network.

• speci_total Specificity for the whole network.

• accu_total Accuracy for the whole network.

• k Number of items.

ghosoya/ludwig documentation built on Sept. 20, 2020, 3:48 p.m.