plot.af {mplot} | R Documentation |
Plot diagnostics for an af object
Description
Summary plot of the bootstrap results of an af object.
Usage
## S3 method for class 'af'
plot(
x,
pch,
interactive = FALSE,
classic = NULL,
tag = NULL,
shiny = FALSE,
best.only = FALSE,
width = 800,
height = 400,
fontSize = 12,
left = 50,
top = 30,
chartWidth = "60%",
chartHeight = "80%",
backgroundColor = "transparent",
legend.position = "top",
model.wrap = NULL,
legend.space = NULL,
options = NULL,
...
)
Arguments
x |
|
pch |
plotting character, i.e., symbol to use |
interactive |
logical. If |
classic |
logical. Depricated. If |
tag |
Default NULL. Name tag of the objects to be extracted from a gvis (googleVis) object. The default tag for is NULL, which will
result in R opening a browser window. Setting |
shiny |
Default FALSE. Set to TRUE when using in a shiny interface. |
best.only |
logical determining whether the output used the
standard fence approach of only considering the best models
that pass the fence ( |
width |
Width of the googleVis chart canvas area, in pixels. Default: 800. |
height |
Height of the googleVis chart canvas area, in pixels. Default: 400. |
fontSize |
font size used in googleVis chart. Default: 12. |
left |
space at left of chart (pixels?). Default: "50". |
top |
space at top of chart (pixels?). Default: "30". |
chartWidth |
googleVis chart area width.
A simple number is a value in pixels;
a string containing a number followed by |
chartHeight |
googleVis chart area height.
A simple number is a value in pixels;
a string containing a number followed by |
backgroundColor |
The background colour for the main area of the chart. A simple HTML color string, for example: 'red' or '#00cc00'. Default: 'transparent' |
legend.position |
legend position, e.g. |
model.wrap |
Optional parameter to split the legend names
if they are too long for classic plots. |
legend.space |
Optional parameter to add additional space between the legend items for the classic plot. |
options |
If you want to specify the full set of googleVis options. |
... |
further arguments (currently unused) |
Details
For each value of c
a parametric
bootstrap is performed under the full model.
For each bootstrap
sample we identify the smallest model inside the fence,
\hat{\alpha}(c)
. We calculate the empirical probability of selecting
model \alpha
for a given value of c
as
p^*(c,\alpha)=P^*\{\hat{\alpha}(c)=\alpha\}.
Hence, if B
bootstrap replications are performed,
p^*(c,\alpha)
is the
proportion of times that model \alpha
is selected. Finally,
define an overall selection probability,
p^*(c)=\max_{\alpha\in\mathcal{A}}p^*(c,\alpha)
and we plot
p^*(c)
against c
. The points on the scatter plot are
colour coded by the model that yielded the highest inclusion probability.