abrasion
These data are the results from an experiment measuring three variables for 30 rubber specimens [Davies, 1957;
Cleveland and Devlin, 1988]: tensile strength, hardness and abrasion loss. The abrasion loss is the amount of material
abraded per unit of energy. Tensile strength measures the force required to break a specimen (per unit area). Hardness
is the rebound height of an indenter dropped onto the specimen. Abrasion loss is measured in g/hp-hour; tensile
strength is measured in kg/cm2; and the unit for hardness is Shore. The intent of the experiment was to determine
the relationship between abrasion loss with respect to hardness and tensile strength.
animal
This data set contains the brain weights and body weights of several types of animals [Crile and Quiring, 1940].
According to biologists, the relationship between these two variables is interesting, since the ratio of brain
weight to body weight is a measure of intelligence [Becker, Cleveland and Wilks, 1987]. The MAT-file contains variables:
AnimalName, BodyWeight, and BrainWeight.
BPM data sets
There are several BPM data sets included with the text. We have iradbpm, ochiaibpm, matchbpm and L1bpm. Each data
file contains the interpoint distance matrix for 503 documents and an array of class labels, as described in Chapter
1. These data can be reduced using ISOMAP before applying other analyses.
calibrat
This data set reflects the relationship between radioactivity counts (counts) to hormone level for 14 immunoassay
calibration values (tsh). The original source of the data is Tiede and Pagano [1979], and we downloaded them from
the website for Simonoff [1996]: www.stern.nyu.edu/SOR/SmoothMeth.
cereal
These data were obtained from ratings of eight brands of cereal [Chakrapani and Ehrenberg, 1981; Venables and Ripley,
1994]. The cereal file contains a matrix where each row corresponds to an observation and each column represents
one of the variables or the percent agreement to statements about the cereal. The statements are: comes back to,
tastes nice, popular with all the family, very easy to digest, nourishing, natural flavor, reasonably priced, a
lot of food value, stays crispy in milk, helps to keep you fit, fun for children to eat. It also contains a cell
array of strings (labs) for the type of cereal.
environmental
This file contains data comprising 111 measurements of four variables. These include Ozone (PPB), SolarRadiation
(Langleys), Temperature (Fahrenheit), and WindSpeed (MPH). These were initially examined in Bruntz, et al. [1974],
where the intent was to study the mechanisms that might lead to air pollution.
ethanol
A single-cylinder engine was run with either ethanol or indolene. This data set contains 110 measurements of compression
ratio (Compression), equivalence ratio (Equivalence), and N0x in the exhaust (NOx). The goal was to understand
how N0x depends on the compression and equivalence ratios [Brinkman, 1981; Cleveland and Devlin, 1988].
example96
This is the data used in Example 9.6 to illustrate the box-percentile plots.
example104
This loads the data used in Examples 10.10 and 10.12 to illustrate parallel coordinate plots. It is a subset of
the BPM data that was reduced using ISOMAP as outlined in Example 10.4.
forearm
These data consist of 140 measurements of the length in inches of the forearm of adult males [Hand, et al., 1994;
Pearson and Lee, 1903].
galaxy
The galaxy data set contains measurements of the velocities of the spiral galaxy NGC 7531. The array EastWest contains
the velocities in the east-west direction, covering around 135 arc sec. The array NorthSouth contains the velocities
in the north-south direction, covering approximately 200 arc sec. The measurements were taken at the Cerro Tololo
Inter-American Observatory in July and October of 1981 [Buta, 1987].
geyser
These data represent the waiting times (in minutes) between eruptions of the Old Faithful geyser at Yellowstone
National Park [Hand, et al., 1994; Scott, 1992].
hamster
This data set contains measurements of organ weights for hamsters with congenital heart failure [Becker and Cleveland,
1991]. The organs are heart, kidney, liver, lung, spleen and testes.
iris
The iris data were collected by Anderson [1935] and were analyzed by Fisher [1936] (and many statisticians since
then!). The data set consists of 150 observations containing four measurements based on the petals and sepals of
three species of iris. The three species are: Iris setosa, Iris virginica, and Iris versicolor. When the iris data
file is loaded, you get three 50 x 4 matrices, one corresponding to each species.
leukemia
The leukemia data set is described in detail in Chapter 1. It measures the gene expression levels of patients with
acute leukemia.
lsiex
This file contains the term-document matrix used in Example 2.3.
lungA, lungB
The lung data set is another one that measures gene expression levels. Here the classes correspond to various types
of lung cancer.
oronsay
The oronsay data set consists of particle size measurements. It is described in Chapter 1. The data can be classified
according to the sampling site as well as the type (beach, dune, midden).
playfair
This data set is described in Cleveland [1993] and Tufte [1983], and it is based on William Playfair’s (1801) published
displays of demographic and economic data. The playfair data set consists of the 22 observations representing the
populations (thousands) of cities at the end of the 1700s and the diameters of the circles Playfair used to encode
the population information. This MAT-file also includes a cell array containing the names of the cities.
pollen
This data set was generated for a data analysis competition at the 1986 Joint Meetings of the American Statistical
Association. It contains 3848 observations, each with five fictitious variables: ridge, nub, crack, weight, and
density. The data contain several interesting features and structures. See Becker, et al. [1986] and Slomka [1986]
for information and results on the analysis of these artificial data.
posse
The posse file contains several data sets generated for simulation studies in Posse [1995b]. These data sets are
called croix (a cross), struct2 (an L-shape), boite (a donut), groupe (four clusters), curve (two curved groups),
and spiral (a spiral). Each data set has 400 observations in 8-D. These data can be used in PPEDA and other data
tours.
salmon
The salmon data set was downloaded from the website for the book by Simonoff [1996]: www.stern.nyu.edu/SOR/SmoothMeth.
The MAT-file contains 28 observations in a 2-D matrix. The first column represents the size (in thousands of fish)
of the annual spawning stock of Sockeye salmon along the Skeena River from 1940 to 1967. The second column represents
the number of new catchable-size fish or recruits, again in thousands of fish.
scurve
This file contains data randomly generated from an S-curve manifold. See Example 3.5 for more information.
singer
This file contains several variables representing the height in inches of singers in the New York Choral Society
[Cleveland, 1993; Chambers, et al., 1983]. There are four voice parts: sopranos, altos, tenors, and basses. The
sopranos and altos are women, and the tenors and basses are men.
skulls
These data were taken from Cox and Cox [2000]. The data originally came from a paper by Fawcett [1901], where they
detailed measurements and statistics of skulls belonging to the Naqada race in Upper Egypt. The skulls file contains
an array called skullsdata for forty observations, 18 of which are female and 22 are male. The variables are greatest
length, breadth, height, auricular height, circumference above the superciliary ridges, sagittal circumference,
cross-circumference, upper face height, nasal breadth, nasal height, cephalic index, and ratio of height to length.
software
This file contains data collected on software inspections. The variables are normalized by the size of the inspection
(the number of pages or SLOC – single lines of code). The file software.mat contains the preparation time in minutes
(prepage, prepsloc), the total work hours in minutes for the meeting (mtgsloc), and the number of defects found
(defpage, defsloc). A more detailed description can be found in Chapter 1.
spam
These data were downloaded from the UCI Machine Learning Repository:
http://www.ics.uci.edu/~mlearn/MLRepository/.html
Anyone who uses email understands the problem of spam, which is unsolicited email, commercial or otherwise. For
example, spam can be chain letters, pornography, advertisements, foreign money-making schemes, etc. This data set
came from Hewlett-Packard Labs and was generated in 1999. The spam data set consists of 58 variables: 57 continuous
and one class label. If an observation is labeled class 1, then it is considered to be spam. If it is of class
0, then it is not considered spam. The first 48 attributes represent the percentage of words in the email that
match some specified word corresponding to spam or not spam. There are an additional six variables that specify
the percentage of characters in the email that match a specified character. Others refer to attributes relating
to uninterrupted sequences of capital letters. More information on the attributes is available at the above internet
link. One can use these data to build classifiers that will discriminate between spam and non-spam emails. In this
application, a low false positive rate (classifying an email as spam when it is not) is very important.
sparrow
These data are taken from Manly [1994]. They represent some measurements taken on sparrows collected after a storm
on February 1, 1898. Eight morphological characteristics and the weight were measured for each bird, five of which
are provided in this data set. These are on female sparrows only. The variables are total length, alar extent,
length of beak and head, length of humerus, and length of keel of sternum. All lengths are in millimeters. The
first 21 of these birds survived, and the rest died.
swissroll
This data file contains a set of variables randomly generated from the Swiss roll manifold with a hole in it. It
also has the data in the reduced space (from ISOMAP and HLLE) that was used in Example 3.5.
votfraud
These data represent the Democratic over Republican pluralities of voting machine and absentee votes for 22 Philadelphia
County elections. The variable machine is the Democratic plurality in machine votes, and the variable absentee
is the Democratic plurality in absentee votes [Simonoff, 1996].
yeast
The yeast data set is described in Chapter 1. It contains the gene expression levels over two cell cycles and five
phases.