sgee (version 0.6-0)

plot.sgee: Coefficient Traceplot Function

Description

Function to produce the coefficent traceplot, with capabilities to account for covariate groups. Used in place of the plot function.

Usage

# S3 method for sgee
plot(x, y, penaltyFun = NULL, main = NULL,
  xlab = "Iterations", ylab = expression(beta), dropIntercept = FALSE,
  trueBeta = NULL, color = TRUE, manualLineColors = NULL,
  pointSpacing = 3, cutOff = NULL, ...)
# S3 method for sgeeSummary
plot(x, y, ...)

Arguments

x

Path of coefficient Estimates.

y

Optional parameter inherited from plot(x,y,...); not used with sgee.

penaltyFun

Optional function that when provided results ina plot of the coefficient estimates verus the corresponding penalty value. When no penaltyFun value is given, the plot generated is of the coefficent estimates versus the iteration number.

main

Optional title of plot.

xlab

Label of x axis; default value is 'Iterations'.

ylab

Label of y axis; default value is the beta symbol.

dropIntercept

Logical parameter indicating whether the intercept estimates should be dropped from the plot (i.e. not plotted). The default is FALSE.

trueBeta

The true coefficient values. If the true coefficient values can be provided, then coefficient estimates that are false positive identifications as non-zero are marked in the plot.

color

Logical parameter indicating that a plot using colors to differentiate coefficients is desired.

manualLineColors

Vector of desired line colors; must match dimension of line colors needed (i.e. same number of colors as there are groups if grouped covariates are sharing a color).

pointSpacing

Space between marks used to indicate a coefficient is a false positive. Spacing is measured in terms of number of indices of the path matrix between marks.

cutOff

Integer value indicating that only the first cutOff steps are to be plotted. Default value is NULL, indicating all steps are to be plotted.

...

Not currently used.

Details

plot.sgee is meant to allow for easy visualization of paths of stagewise (or regularized) coefficient estimates. A great deal of flexibility is provided in terms of how the plot is presented. The poenaltyFun paramter allows for a penalty function to be provided (such as the $L_1$ norm) to plot the coefficietn estimates against. When given the trueBeta parameter, the plot marks the paths of coefficient estimates that are falsely identified as being non zero. Finally, a switch for black and white versus color plots is provided (color).

Examples

# NOT RUN {

#####################
## Generate test data
#####################

## Initialize covariate values
p <- 50 
beta <- c(rep(2.4,5),
          c(1.3, 0, 1.7, 0, .5),
          rep(0.5,5),
          rep(0,p-15))
groupSize <- 1
numGroups <- length(beta)/groupSize



generatedData <- genData(numClusters = 50,
                         clusterSize = 4,
                         clusterRho = 0.6,
                         clusterCorstr = "exchangeable",
                         yVariance = 1,
                         xVariance = 1,
                         numGroups = numGroups,
                         groupSize = groupSize,
                         groupRho = 0.3,
                         beta = beta,
                         family = gaussian(),
                         intercept = 0)

genDF <- data.frame(generatedData$y, generatedData$x)
coefMat <- bisee(formula(genDF),
                 data = genDF,
                 lambda1 = 0,         ##effectively see
                 lambda2 = 1,
                 family = gaussian(),
                 clusterID = generatedData$clusterID, 
                 corstr="exchangeable", 
                 maxIt = 200,
                 epsilon = .1)
############################
## Various options for plots
############################

par(mfrow = c(2,2))

## plain useage
plot(coefMat, main = "Plain Usage")

## With penalty
plot(coefMat, penaltyFun = function(x){sum(abs(x))}, xlab
= expression(abs(abs(beta))[1]), main = "With Penalty")

## using true beta value to highlight misclassifications
plot(coefMat, trueBeta = beta, main = "ID Missclassification")

## black and white option
plot(coefMat, trueBeta = beta, color = FALSE, main =
"Black and White", pointSpacing = 5)



# }