R2GUESS (version 2.0)

plotMPPI: Plots the marginal posterior probability of inclusion (MPPI) for each predictor

Description

The plotMPPI function plots the marginal posterior probability of inclusion (MPPI) of each predictor.

Usage

plotMPPI(x, threshold.model = 0.01,
    threshold.variable = 0.1, Figure = TRUE, cutoff = TRUE, useMC = FALSE)

Arguments

x

an object of class ESS.

threshold.model

either an integer representing the number of model to be retained in the list of best models, or a value defining the minimal model posterior probability for inclusion.

threshold.variable

threshold probability for selecting the most relevant predictors. This threshold can be calibrated by controlling the FDR using FDR.permutation.

Figure

if TRUE (by default) will generate the MPPI plot. If FALSE only information on the selected predictors will provided.

cutoff

if TRUE (by default) will plot an horizontal line representing the cut-off value indicating by the argument threshold.variable. If FALSE the cut-off value is not plotted.

useMC

if TRUE, use simple Monte Carlo estimation for the MPPI across all visited models.

Value

The plotMPPI function returns information on the best models (i.e. those satisfying the threshold.model criterion) and on the most relevant predictors. (above threshold.variable).

Rank

the rank on the models selected.

nVisits

number of times each model has been visited along the run.

ModSize

number of predictors in each of the best models.

logCondPost

the log conditional posterior for each model.

Jeffries

Jeffrie's scale value for each model.

postProb

posterior probability of each model.

modelName

list of predictors in each of the best models.

modelPosInX

position (in the predictor matrix) of the constituents of the best models.

var.TOP.MPI

predictors with MPPI>threshold.variable and belonging to the best models.

var.MPI

predictors which have a MPPI greater than threshold.variable.

Examples

# NOT RUN {
modelY_Hopx <- example.as.ESS.object()
# To get a large plot 
# dev.new(width=13,height=6)
MPPI.Hopx <- plotMPPI(modelY_Hopx,threshold.model=20,threshold.variable=0.45)
print(MPPI.Hopx)
# }