GGMfit: R6 Class for estimation of Gaussian Graphical Model

In cambroise/LITree: Latent Incomplete Tree Inference via Graphical Gaussian Model

Description

The Class GGMfit estimate a Gaussian Graphical Model. It can use 5 different models and associated estimation procedure with or without bagging. Three of the algorithms (em.latent.trees, em.glasso, em.chow.liu) take into account missing variables while the two other do not (glasso, chow.liu).

Usage

experiment = GGMfit$new(X,method="glasso",nb.missing.var=0,fit.number=20,...)

Arguments

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

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

- method The name of the estimation procedure (em.latent.trees, em.glasso, em.chow.liu, glasso, chow.liu)

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

- K.score Array of prediction of edges. The array is 2 dimensional homogeneous in dimension to the adjacency matrix of the graph to be inferred.

Methods

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

$run(bagging=FALSE,nb.bootstrap=29) Running the experiment (with or without using bagging)

Examples

## Not run: 
star.graph <- graphModel$new(type = "starerdos",size=30, p.or.m = 0.05)
star.model <- GGMmodel$new(graph=star.graph)
star.model.missing <- GGMmodel$new(graph=star.graph,nb.missing.var= 1)
dim(star.model.missing$getAdjmat())
dim(star.model.missing$getAdjmatCond())
star.model.missing$randomSample(n=60)
dim(star.model.missing$getX())
dim(star.model.missing$getXobs())
star.model.missing.fit <-GGMfit$new(star.model.missing$getXobs(),fit.number = 20,method="glasso")
star.model.missing.fit$run()
star.model.missing.fit2 <-GGMfit$new(star.model.missing$getXobs(),nb.missing.var= 1,fit.number = 20,method="em.latent.trees")
star.model.missing.fit2$run()

## End(Not run)

cambroise/LITree documentation built on May 6, 2019, 8:32 p.m.