kibler: 1985 Auto Imports Database
In rrcovHD: Robust Multivariate Methods for High Dimensional Data

Description Usage Format Details Source References Examples

The original data set kibler.orig consists of three types of entities: (a) the specification of an auto in terms of various characteristics, (b) its assigned insurance risk rating and (c) its normalized losses in use as compared to other cars.

The second rating corresponds to the degree to which the auto is more risky than its price indicates. Cars are initially assigned a risk factor symbol associated with its price. Then, if it is more risky (or less), this symbol is adjusted by moving it up (or down) the scale. Actuarians call this process "symboling". A value of +3 indicates that the auto is risky, -3 that it is probably pretty safe.

The third factor is the relative average loss payment per insured vehicle year. This value is normalized for all autos within a particular size classification (two-door small, station wagons, sports/speciality, etc...), and represents the average loss per car per year.

1	data(kibler)

A data frame with 195 observations on the following 14 variables. The original data set (also available as kibler.orig) contains 205 cases and 26 variables of which 15 continuous, 1 integer and 10 nominal. The non-numeric variables and variables with many missing values were removed. Cases with missing values were removed too.

symboling: a numeric vector
wheel-base: a numeric vector
length: a numeric vector
width: a numeric vector
height: a numeric vector
curb-weight: a numeric vector
bore: a numeric vector
stroke: a numeric vector
compression-ratio: a numeric vector
horsepower: a numeric vector
peak-rpm: a numeric vector
city-mpg: a numeric vector
highway-mpg: a numeric vector
price: a numeric vector

The original data set contains 205 cases and 26 variables of which 15 continuous, 1 integer and 10 nominal. The non-numeric variables and variables with many missing values were removed. Cases with missing values were removed too. Thus the data set remains with 195 cases and 14 variables.

www.cs.umb.edu/~rickb/files/UCI/

Kibler, D., Aha, D.W. and Albert, M. (1989). Instance-based prediction of real-valued attributes. Computational Intelligence, Vo.l 5, 51-57.

data(kibler)
x.sd <- apply(kibler,2,sd)
xsd <- sweep(kibler, 2, x.sd, "/", check.margin = FALSE)
apply(xsd, 2, sd)

x.mad <- apply(kibler, 2, mad)
xmad <- sweep(kibler, 2, x.mad, "/", check.margin = FALSE)
apply(xmad, 2, mad)

x.qn <- apply(kibler, 2, Qn)
xqn <- sweep(kibler, 2, x.qn, "/", check.margin = FALSE)
apply(xqn, 2, Qn)


## Display the scree plot of the classical and robust PCA
screeplot(PcaClassic(xsd))
screeplot(PcaGrid(xqn))

#########################################
##
## DD-plots
##
## Not run: 
usr <- par(mfrow=c(2,2))
plot(SPcaGrid(xsd, lambda=0, method="sd", k=4), main="Standard PCA")    # standard
plot(SPcaGrid(xqn, lambda=0, method="Qn", k=4))                         # robust, non-sparse

plot(SPcaGrid(xqn, lambda=1,43, method="sd", k=4), main="Stdandard sparse PCA")  # sparse
plot(SPcaGrid(xqn, lambda=2.36, method="Qn", k=4), main="Robust sparse PCA")     # robust sparse
par(usr)

#########################################
##  Table 2 in Croux et al
##  - to compute EV=Explained variance and Cumulative EV we
##      need to get all 14 eigenvalues
##
rpca <- SPcaGrid(xqn, lambda=0, k=14)
srpca <- SPcaGrid(xqn, lambda=2.36, k=14)
tab <- cbind(round(getLoadings(rpca)[,1:4], 2), round(getLoadings(srpca)[,1:4], 2))

vars1 <- getEigenvalues(rpca);  vars1 <- vars1/sum(vars1)
vars2 <- getEigenvalues(srpca); vars2 <- vars2/sum(vars2)
cvars1 <- cumsum(vars1)
cvars2 <- cumsum(vars2)
ev <- round(c(vars1[1:4], vars2[1:4]),2)
cev <- round(c(cvars1[1:4], cvars2[1:4]),2)
rbind(tab, ev, cev)

## End(Not run)