Graphics Reference
In-Depth Information
points within each site are quite linear. Panel b again employs different symbols to
indicate leaf node membership according to the piecewise best simple linear model
inFig.
.
.Wesee that nodemembership divides the data into three clusters,with the
first cluster belonging to Sites
and
and the second and third clusters to Sites
,
,
and
, depending on whether or not Depth
.
. he first cluster (indicated by cir-
cles)clearlyexhibitsthemostpronouncedcurvature.hissuggeststhatthenonlinear
relationship between Muscle and SIR
is mainly due to the observations from Sites
and
. On the other hand, we saw in Fig.
.
a that at these two sites, Muscle varies
roughly linearly with Shell. hus it is likely that the curvature in Fig.
.
is at least
partly due to the power transformation of Shell in the definition of SIR
in Eq. (
.
).
Crash Tests - Outlier Detection
Under Confounding
7.5
he data in this example are obtained from
vehicle crash tests performed by
theNational HighwayTransportation SafetyAdministration (NHTSA)between
and
(http://www-nrd.nhtsa.dot.gov).heresponsevariableisthesquarerootof
the head injury criterion (
hic
) measured on a crash dummy. Values of
hic
range
from
to
,with
beingtheapproximatelevelbeyondwhichapersonisexpected
tosuffersevere headinjury. Twenty-five predictorvariables, definedinTable
.
,pro-
vide information on the vehicles, dummies, and crash tests. Angular variables are
measured clockwise, with
,
, and
degrees corresponding to the driver's let,
front,andrightsides,respectively. Aboutonequarter ofthevehicle modelsaretested
morethan once,with themostotentestedbeing the
Chevy Citation, whichwas
tested
times.
−
Table
.
.
Variables for NHTSA data
Name
Description
Name
Description
hic
Head injury criterion
make
Car manufacturer (
values)
Car model year
mkmodel
Car model (
values)
year
Car bodytype (
values)
Transmission type (
values)
body
transm
Engine type (
values)
Engine displacement(liters)
engine
engdsp
vehtwt
Vehicletotal weight (kg)
colmec
Collapse mechanism (
values)
Vehiclewidth (mm)
Car modification indicator(
values)
vehwid
modind
Vehiclespeed (km/h)
Crabbed angle (degrees)
vehspd
crbang
Track surface (
values)
Principal direction of force (degrees)
tksurf
pdof
Track condition (
values)
Impact angle (degrees)
tkcond
impang
Occupant type (
values)
Dummy size (
values)
occtyp
dumsiz
Seat position (
values)
Barrier rigidity (rigid/deformable)
seposn
barrig
barshp
Barrier shape (
values)
belts
Seat belt type (none/
pt/
pt)
Airbag present (yes/no)
Knee restraint present (yes/no)
airbag
knee