boot
ObjectsThe Boot
function in the car package uses the boot
function from the
boot package to do a straightforward case
or residual bootstrap for many regression objects. These are method functions for standard generics to summarize the results of the bootstrap. Other tools for this purpose are available in the boot
package.
# S3 method for boot
hist(x, parm, layout = NULL, ask, main = "", freq = FALSE,
estPoint = TRUE, point.col = carPalette()[1], point.lty = 2, point.lwd = 2,
estDensity = !freq, den.col = carPalette()[2], den.lty = 1, den.lwd = 2,
estNormal = !freq, nor.col = carPalette()[3], nor.lty = 2, nor.lwd = 2,
ci = c("bca", "none", "perc", "norm"), level = 0.95,
legend = c("top", "none", "separate"), box = TRUE, ...)# S3 method for boot
summary(object, parm, high.moments = FALSE, extremes = FALSE, ...)
# S3 method for boot
confint(object, parm, level = 0.95, type = c("bca", "norm",
"basic", "perc"), ...)
# S3 method for boot
Confint(object, parm, level = 0.95, type = c("bca", "norm",
"basic", "perc"), ...)
# S3 method for boot
vcov(object, ...)
An object created by a call to boot
in the boot
package, or to Boot
in the car package of class "boot"
.
A vector of numbers or coefficient names giving the coefficients for which a histogram or confidence interval is desired. If numbers are used, 1 corresponds to the intercept, if any. The default is all coefficients.
If set to a value like c(1, 1)
or c(4, 3)
, the layout
of the graph will have this many rows and columns. If not set, the program
will select an appropriate layout. If the number of graphs exceed nine, you
must select the layout yourself, or you will get a maximum of nine per page.
If layout=NA
, the function does not set the layout and the user can
use the par
function to control the layout, for example to have
plots from two models in the same graphics window.
If TRUE
, ask the user before drawing the next plot; if FALSE
, don't
ask.
Main title for the graphs. The default is main=""
for no title.
The default for the generic hist
function is freq=TRUE
to give a frequency histogram. The default for hist.boot
is freq=FALSE
to give a density histogram. A density estimate and/or a fitted normal density can be added to the graph if freq=FALSE
but not if freq=TRUE
.
If estPoint=TRUE
, the default, a vertical line is drawn on the histgram
at the value of the point estimate computed from the complete data. The
remaining three optional arguments set the color, line type and line width
of the line that is drawn.
If estDensity=TRUE
andfreq=FALSE
, the default, a kernel density estimate is drawn
on the plot with a call to the density
function with no additional
arguments. The
remaining three optional arguments set the color, line type and line width
of the lines that are drawn.
If estNormal=TRUE
andfreq=FALSE
, the default, a normal density
with mean and sd computed from the data is drawn on the plot. The
remaining three optional arguments set the color, line type and line width
of the lines that are drawn.
A confidence interval based on the bootstrap will be added to the histogram
using the BCa method if ci="bca"
the percentile method if
ci="perc"
, or the normal method if ci="norm"
. No interval is drawn if
ci="none"
. The default is "bca"
. The interval is indicated
by a thick horizontal line at y=0
. For some bootstraps the BCa method is unavailable, in which case a warning is issued and ci="perc"
is substituted. If you wish to see all the options at once, see boot.ci
. The normal method is computed as the (estimate from the original data) minus the bootstrap bias plus or minus the standard deviation of the bootstrap replicates times the appropriate quantile of the standard normal distribution.
A legend can be added to the (array of) histograms. The value “"top"” puts at the top-left of the plots. The value “"separate"” puts the legend in its own graph following all the histograms. The value “"none"” suppresses the legend.
Add a box around each histogram.
Additional arguments passed to hist
; for other methods this is included for compatibility with the generic method. For example, the argument
border=par()$bg
in hist
will draw the histogram transparently, leaving only
the density estimates. With the vcov
function, the additional arguments are passed to cov
. See the Value section, below.
Should the skewness and kurtosis be included in the summary? Default is FALSE.
Should the minimum, maximum and range be included in the summary? Default is FALSE.
Confidence level, a number between 0 and 1. In confint
, level
can be
a vector; for example level=c(.50, .90, .95)
will return the following estimated quantiles: c(.025, .05, .25, .75, .95, .975)
.
Selects the confidence interval type. The types
implemented are the "percentile"
method, which uses the function
quantile
to return the appropriate quantiles for the confidence
limit specified, the default bca
which uses the bias-corrected and accelerated
method presented by Efron and Tibshirani (1993, Chapter 14). For the
other types, see the documentation for boot
.
hist
is used for the side-effect of drawing an array of historgams of
each column of the first argument. summary
returns a matrix of
summary statistics for each of the columns in the bootstrap object. The
confint
method returns confidence intervals. Confint
appends the estimates based on the original fitted model to the left of the confidence intervals.
The vcov
returns the sample covariance of the bootstrap sample estimates. If any of the bootstrap replications returned an NA
value, then vcov
will return a matrix of NA
s. You can compute the covariance of the complete cases by adding the argument use="complete.obs"
to the call to vcov
; see cov
.
Efron, B. and Tibsharini, R. (1993) An Introduction to the Bootstrap. New York: Chapman and Hall.
Fox, J. and Weisberg, S. (2019) An R Companion to Applied Regression, Third Edition. Thousand Oaks: Sage.
Fox, J. and Weisberg, S. (2017) Bootstrapping, http://socserv.mcmaster.ca/jfox/Books/Companion/appendix/Appendix-Bootstrapping.pdf.
Weisberg, S. (2013) Applied Linear Regression, Fourth Edition, Wiley
See Also Boot
, hist
,
density
, Fox and Weisberg (2017), cited above
# NOT RUN {
m1 <- lm(Fertility ~ ., swiss)
betahat.boot <- Boot(m1, R=99) # 99 bootstrap samples--too small to be useful
summary(betahat.boot) # default summary
confint(betahat.boot)
hist(betahat.boot)
# }
Run the code above in your browser using DataLab