plotCurveEstimate: Generic function for plotting a fractional polynomial curve estimate

Description

Plot a fractional polynomial curve estimate for either a single model or a Bayesian model average over BayesMfp objects. Optionally, credible intervals and / or bands can be added to the plot.

Usage

plotCurveEstimate(model, termName, plevel = 0.95, slevel = plevel,
plot = TRUE, legendPos = "topleft", rug = FALSE, partialResids=TRUE, 
hpd=TRUE,..., main = NULL)

Value

a list of various plotting information:

original: grid on the original covariate scale
grid: grid on the transformed scale
mode: mode curve values, only for BayesMfp object
mean: pointwise mean curve values, only for BmaSamples object
median: pointwise median curve values, only for BmaSamples object
plower: lower boundaries for pointwise intervals
pupper: upper boundaries for pointwise intervals
slower: lower boundaries for SCB
supper: upper boundaries for SCB
obsVals: observed values of the covariate on the original scale
sampleSize: sample size underlying the curve estimate, only for BmaSamples object
partialResids: partial residuals
transform: vector of shift and scale parameter

Arguments

model: an object of class BayesMfp or BmaSamples
termName: string denoting an FP term, as written by the summary method
plevel: credible level for pointwise intervals, and NULL means no pointwise intervals (default: 0.95)
slevel: credible level for simultaneous credible band (SCB), NULL means no SCB (defaults to plevel)
plot: if FALSE, only return values needed to produce the plot, but do not plot (default is TRUE, so a plot is made)
legendPos: position of coefficient estimates (for BayesMfp) or sample size (for BmaSamples) in the plot, NULL suppresses the printing (default is “topleft”)
rug: add a rug to the plot? (default: FALSE)
partialResids: add partial residuals to the plot? (default: TRUE)
hpd: use HPD intervals (TRUE, default) or quantile-based (FALSE) intervals?
...: further arguments in case of a BayesMfp object (see details) and arguments for plotting with matplot
main: optional main argument for the plot

Details

Further arguments for application on a BayesMfp object:

grid: vector of unscaled abscissae, default is a length gridSize grid over the observed range specified by providing the argument NULL.
post: list with posterior parameters of the model, which may be provided manually to accelerate plotting in a loop
gridSize: default number of grid points used when no grid is supplied (default is 201)
numSim: number of simulations for estimation of the SCB (default is 500)

Examples

Run this code

## construct a BayesMfp object
set.seed(19)
x1 <- rnorm (n=15)
x2 <- rbinom (n=15, size=20, prob=0.5) 
x3 <- rexp (n=15)
y <- rt (n=15, df=2)

test <- BayesMfp (y ~ bfp (x2, max = 4) + uc (x1 + x3), nModels = 100,
method="exhaustive")

## plot the x2 curve estimate for the 20-th best model
p1 <- plotCurveEstimate (test[20], "x2")

## look at the returned list
str(p1)

## plot the BMA curve estimate for the same covariate
testBma <- BmaSamples (test)
p2 <- plotCurveEstimate (testBma, "x2")

## look at the returned list
str(p2)

## try the new options:
plotCurveEstimate (testBma, "x2", partialResids=FALSE, hpd=FALSE)