GGMexperiment: R6 Class for running experiment of Gaussian Graphical Model
In cambroise/LITree: Latent Incomplete Tree Inference via Graphical Gaussian Model

Description Usage Arguments Methods Examples

The class aims at comparing multiple Gaussian Graphical Model Inference procedure when a ground truth is available

1	experiment = GGMexperiment$new(X.list, adjmat, methods=c("glasso","em.latent.trees"))

- X.list A list of data.frame, which will be used as input data for the inference

- adjmat The ground truth adjacency matrix used for the evaluation (roc curve)

- nb.missing.var Number of missing variable (0 by default)

- methods A vector of characters list the methods to be tested (see GGMfit for the list of possible methods)

- fit.number Number of evaluation point (20 by default)

- nb.sample Number of data frame in X.list (set automatically while initializing the object)

- K.score.array Array of prediction of edges. The array is 3 dimensional (2 first dimension for storing a results, third dimension for compiling all results)

- prediction A dataframe with 3 columns (prediction, label and method) used for methods evaluation

$new(X.list=NULL,adjmat=NULL,nb.missing.var=0,methods="glasso",fit.number=20) Initialize the experiment

$run(bagging=FALSE) Running the experiment (with or without using bagging)

$roc.plot() Plot the roc curves (all methods on the same plot)

## Not run: 
star.graph <- graphModel$new(type = "starerdos",size=30, p.or.m = 0.05)
star.model <- GGMmodel$new(graph=star.graph)
plot(star.model)
star.model$randomSample(n=50)
testingGlasso<-GGMexperiment$new(X.list = list(star.model$getX()), adjmat = star.model$getAdjmat())
testingGlasso$run()
print(glasso.auc<-testingGlasso$auc())
testingGlasso$roc.plot()

## End(Not run)