OS/2 Shareware BBS: Science/Science.zip/diamdemo.zip/tutorial.doc

Text File | 1993-12-08 | 37KB | 1,018 lines

DEMO TUTORIAL

This tutorial leads you through a simple exploration of sample
data, using some of the features of BMDP/DIAMOND. It is assumed
that you know how to use a mouse with window applications. You
should read this while actually working with BMDP/DIAMOND at a
workstation or computer. The tutorial will take approximately
thirty minutes to complete.

In this tutorial you will learn how to:

*    open a dataset
*    view the Directory window
*    get HELP
*    open Pair-Wise windows
*    color subsets of data points
*    show best-fit straight lines
*    exit BMDP/DIAMOND
*    obtain univariate statistics
*    display bivariate statistics
*    transform variables
*    manipulate graphic representations of data
*    re-invoke BMDP/DIAMOND on a subset of data
*    examine the raw data
*    open various windows in BMDP/DIAMOND
*    learn and use Ice

Before Getting Started

BMDP/DIAMOND has been designed to allow you to quickly and easily
explore and visualize your data. The interface is intuitive and
very flexible. You execute tasks in BMDP/DIAMOND with a combination
of mouse clicks and keystrokes.

Positioning the Cursor

The first step in telling BMDP/DIAMOND what you want it to do is to
position the cursor. The cursor location within a window tells the
program which elements of the display are to be affected.

You can position the cursor within the following areas:

*    In the "black space" of the window, the area outside of any
     plots or numbers in that window
*    Over text in a window
*    Over numbers in a window
*    Over a histogram to operate on one variable
*    Over a scatterplot to operate on two variables
*    Within a plot to manipulate a color region

After you position the cursor, you must press a key or use a mouse
click to tell the program to carry out the command. You can find a
summary of the keystrokes and what they do in the HELP shortcut
keys reference window.

You can open the HELP Shortcut Keys Reference window at any time by
positioning the cursor in the black space of almost any window and
pressing the question mark ? key.

You can also operate BMDP/DIAMOND through menus, which are
available in the HELP system.

With a little practice, positioning the cursor in the proper
location will become automatic. But if at first the program seems
to do something unexpected, check to see where you have the cursor
positioned.

In the following examples we will always tell you where to position
the cursor. By the time you work through all of the examples in
this section, the operation of BMDP/DIAMOND will seem perfectly
natural to you.

Starting BMDP/DIAMOND

NOTE: Ensure that the subdirectory where you installed BMDP/DIAMOND
is in your path.

1.   From your current BMDP/DIAMOND working directory type the
command:

     bmdpdiam

2.   Press Enter.

NOTE: You can also double click on the bmdpdiam icon in the file
manager.

The system displays the BMDP/DIAMOND Logo window.

3.   Click on OK in the upper left hand corner.

The system displays the BMDP/DIAMOND Startup window.

4.    Click on econ from the list of sample dataset names to choose
a dataset.

The econ data shows 98 quarters from 1967 to 1991 and includes the
value of 12 economic variables for each quarter.

The system displays the Directory window for the dataset econ.

NOTE: If you have trouble starting BMDP/DIAMOND, refer to the
installation procedures.

THE DIRECTORY WINDOW

The Directory window is a concise, global, pictorial view of all
the displayed variables. You can access all other windows from the
Directory window. The elements of the Directory window are shown in
the following figure.

Histograms and cumulative histograms for each variable appear in
the columns along the left. The mirror image of each histogram and
cumulative histogram appears in rows along the bottom of the
Directory window. Variable names are displayed to the right of the
histogram columns, and above the histogram rows. Scatterplots of
all pair-wise combinations of variables appear in a matrix to the
right of the histogram columns.

ON-LINE HELP

Context-sensitive on-line HELP is available to provide
descriptions, explanations, and menu alternatives for performing
all operations.

To access HELP:

1.   Click on HELP located in the upper left hand corner of the
Directory window.

The system displays the BMDP/DIAMOND Help window.

2.   Click on the Describe BMDP/DIAMOND option at the top of the
menu.

The system displays the Describe BMDP/DIAMOND Help window.

NOTE:     HELP contains nested windows. Pressing Esc within any
HELP window will close the window and all its descendants.

3.   Place the cursor within the Describe BMDP/DIAMOND window, and
press the Esc key to close the window.

To access short cut keys:

The HELP facility also provides a list and explanation of shortcut
keys.

1.   Click on Open the Shortcut Keys Reference window option in the
Using the Main Directory window.

he system displays the Shortcut Keys Reference window.

The Quick Reference Card contained in your BMDP/DIAMOND Demo
package provides all of the information in this window, as well as
other useful information. Both the Shortcut window and the card
will be valuable tools to aid you in becoming familiar with
BMDP/DIAMOND.

2.   Place the cursor over the window and press Esc to close the
Shortcut Keys Reference window.

You can access Help from most windows in BMDP/DIAMOND.

WORKING WITH WINDOWS

Working with BMDP/DIAMOND windows is similar to working with other
windows  applications you may be using with your operating system.
You can maximize, enlarge, or move windows in BMDP/DIAMOND. (See
your operating system documentation for details on working with
windows.)

PAIR-WISE WINDOW

You can use the Pair-Wise window to display a more detailed view of
the relationship between any pair of variables.

To access the Pair-Wise window for DJI/Year:

1.   Place the cursor on the DJI/Year scatterplot (where DJI on the
left column intersects with Year on the bottom), and press the
Enter key.

The system displays the Pair-Wise window.

The Pair-Wise window contains a view of the histograms, cumulative
histograms, and the scatterplot for the pair of variables that you
selected.

2.   Place the cursor on the title bar of the Pair-Wise window and
hold down the left button.

3.   Drag the window to the upper right-hand corner of the screen
so that you can still see most of the Directory window and release
the mouse button.

4.   Move the cursor into the Directory window.

Notice that when you move the cursor from the Pair-Wise window to
the Directory window, the title bar automatically changes color.
This indicates that the window is active and, although not in
front, is ready to accept your input.

When you want more information on this window, click on HELP from
the Pair-Wise window.

COLORING SUBSETS OF DATA

You can color code data points in many different ways in
BMDP/DIAMOND. For example, you can set a rainbow of colors to cover
all or part of the range of a variable. BMDP/DIAMOND also permits
you to assign specific colors to subsets of your data. You can
independently assign red, blue, and green to the data points of
your choice. Red, blue, and green combine to make yellow, cyan,
magenta, and white.

Applying a Large Rainbow of Colors

1.   In the Directory window, place the cursor over the top left
histogram for the variable Term, and press the exclamation point !
key.

The system displays a rainbow of colors covering a large range of
the selected variable. Notice that all of the histograms and
scatterplots reflect the color coding.

2.   To remove all of the colors, move the cursor within the black
space of the Directory window, and press the Home key.

NOTE: You can also remove the colors by positioning the cursor
within the plot where you set the colors and pressing the Home key.

Applying a Small Rainbow of Colors.

1.   Position the cursor within the center of the Year histogram in
the DJI/Year Pair-Wise window, and press the asterisk * key.

The system displays a rainbow covering a small range of the
selected variable.

2.   Press the Home key within any one of the plots in the Pair-
Wise window to remove the colors.

Applying Specific Colors to Subsets of your Data

You can use the first letter of the name of the color to mark your
data. This option is handy for identifying outliers or other points
of interest. You can select specific points or blocks of data in
the scatterplots directly, or you can select blocks of data using
the histograms.

To set the lower boundary for a color block:

To color the years when the Dow Jones Industrial average was low
(before the sharp rise around 1980), start by setting the lower
boundary

1.   In the DJI/Year Pair-Wise window, place the cursor inside of
the Year histogram near the left edge, and press the r (lower case)
key.

This action sets the lower boundary.

Selecting the lower case letter of the color sets the lower
boundary. All cases above the year selected by your cursor are now
colored red in every histogram and scatterplot. The upper boundary
defaults to the greatest value of the variable.

To set the upper boundary for a color block:

1.   In the DJI/Year Pair-Wise window, place the cursor inside the
Year histogram at its midpoint, and press the R (upper case) key.

This action sets the upper boundary.

The red coloring is now restricted to cases in the years prior to
the steep rise in the stock market around 1980.

Using the mouse to adjust the boundaries:

You can also use the mouse to adjust the lower and upper
boundaries.

1.   In the DJI/Year Pair-Wise window, place the cursor inside the
red area near the left edge of the Year histogram and hold down the
left button.

2.   Drag the mouse to the edge (or beyond) of the left side of the
histogram and release the button.

The system extends the red area to the left edge of the histogram.

NOTE: Changes in color assignment in the histogram are reflected in
all plots in all open windows.

Using colors to select subsets of your data is a valuable utility
in BMDP/DIAMOND. You have many other options available in color
selection. To learn more about your color selection options,
consult the documentation or click on HELP from the Directory
window.

BEST-FIT STRAIGHT LINES

You can draw simple regression lines in the scatterplots of Pair-
Wise windows. To draw simple regression lines:

1.   In the DJI/Year Pair-Wise window, place the cursor within the
black space of the Directory window, and press the equal sign  =
key.

The system displays the Pair-Wise window with the simple regression
lines.

The system displays one line for each existing color group in its
own color, a gray line for the uncolored points, and a white line
for all the points together. If you change the ranges of the
colored regions, BMDP/DIAMOND automatically computes and displays
the correct  regression lines.

To remove the simple regression lines:

1.   Press the = key again. When you want more information on
regression lines, click on HELP from the Pair-Wise window.

RAW DATA WINDOW

To examine the raw data:

1.   Place the cursor within the black space of the Directory
window, and press the asterisk * key.

The system displays the Raw Data window for the dataset econ. You
can use this window to review the data.

2.   To scroll the data either vertically or horizontally, place
the cursor in the Raw Data window and use the arrow keys.

3.   Place the cursor within the Raw Data window, and press the Esc
key to close the window.

When you want more information on this window, click on HELP from
this window.

UNIVARIATE STATISTICS WINDOW

To bring up the Univariate Statistics window:

1.   Place the cursor within the black space of the Directory
window, and press the ampersand & key.

BMDP/DIAMOND displays eleven different statistics that you can
manipulate and sort.

NOTE: The Univariate Statistics window may be too wide to see in a
single view if you have a small screen. If you have that problem,
move the window right and left by dragging the title bar of the
window.

2.   Place the cursor in the title bar of the Univariate Statistics
window, hold down the left mouse button, and drag the window to the
left to view more of the statistics.

In general, in many of BMDP/DIAMOND windows, you can view selected
subsets of data by pressing the first letter of the color.

3.   Place the cursor in the Univariate Statistics window, and
press the r key to view the univariate statistics for the red
cases.

4.   Press the n key to view the uncolored cases.

5.   Press the a key to view all of the points.

6.   Place the cursor within the Univariate Statistics window and
press the Esc key to close this window.

When you want more information on this window, click on HELP from
this window.

BIVARIATE STATISTICS WINDOW

BMDP/DIAMOND can display up to five different windows of bivariate
statistics.

To obtain the correlation coefficients:

1.   Place the cursor within the black space of the Directory
window, and press the 1 (numeric) key.

The system displays the Bivariate Correlation window for econ.

NOTE: Statistics are colored according to the color coding of the
data case subsets.

NOTE: The Bivariate Statistics window may be too wide to see in a
single view if you have a small screen. If you have that problem,
move the window right and left by dragging the title bar of the
window.

2.   Place the cursor in the title bar of the Bivariate Statistics
window, and hold down the left mouse button. Drag the window to the
left, and release the mouse button.

This action allows you to view more of the statistics in the
window.

To produce other bivariate statistics, you can type the numbers 0
through 4 within the black space of the Directory window.

3.   To close this window, place the cursor in the Bivariate
Statistics window and press the Esc key.

For more information on these windows, click on HELP from any
Bivariate Statistics window.

TRANSFORMATIONS

You can independently apply a variety of scaling and transformation
functions to the variables.

To apply a natural log scaling function to the DJI variable:

1.   Place the cursor within a DJI histogram in the Directory
window, and press the l (lower case) key.

The system displays the transformed DJI histogram.

Notice that the variable name has changed to ln DJI, and all
histograms and scatterplots reflect the transformation.

To remove all scaling and transform functions, apply the identity
function:

1.   Place the cursor within the black space of the Directory
window or within the DJI histogram, and press the i (lower case)
key.

For more information, click on HELP from the Directory window.

TRIPLE-WISE WINDOWS

To observe a scatterplot of three variables in a three dimensional
presentation, invoke a Triple-Wise window.

To show Unemployment as a function of GNP over Time:

1.   Place the cursor within any Unemployment histogram and press
the back quote ` key.

The cursor changes to a question mark indicating that the system is
waiting for you to make another selection.

2.   Place the cursor within any GNP histogram and press the back
quote ` key.

This action selects the second variable.

3.   Place the cursor within any Year histogram and press the back
quote ` key.

This action selects the third variable. The system displays the
Triple-Wise window for Unemployment/GNP/Year.

To rotate the Triple-Wise window:

1.   Move the cursor into the Triple-Wise window and use the arrow
keys or the less than and greater than keys to rotate the plot.

2.   Press the Home key to return the plot to its original
orientation.

To display subsets of this plot:

1.   Press the r key to view the red cases, the quarters in this
three-way relationship when DJI was low.

2.   Press the n key to view the uncolored cases.

3.   Press the a key to view all of the cases.

4.   Place the cursor within the Triple-Wise window, and press the
Esc key to close the window. For more information on this window,
activate HELP from this window.

PARAMETRIC SNAKE WINDOW

The Parametric Snake window is another three dimensional
presentation. It connects the points of a two-way scatterplot in
order of a third variable. The following example uses Unemployment
as a function of the composite Economic indicator in order of Year
(by quarter).

To access the Parametric Snake Window:

1.   Place the cursor within any Unemployment histogram in the
Directory window and press the tilde ~ key.

The cursor changes to a question mark indicating that the system is
waiting for you to make another selection.

2.   To select the second variable, place the cursor within any
Economy histogram and press the tilde ~ key.

The cursor remains a question mark indicating that the system is
waiting for you to make another selection.

3.   To select the third variable, place the cursor within any Year
histogram and press the tilde ~ key.

The system displays the Snake window for Unemployment/Economy/Year.
The graphic representation vividly demonstrates economic loops that
occur over time.

To display subsets of this plot:

1.   Place the cursor within the black space of the Snake window.

2.   Press the r key to view the red colored cases.

3.   Press the n key to view the uncolored cases.

4.   Press the a key to view all of the cases.

5.   To close this window, place the cursor within the Snake window
and press the Esc key. For more information on this window,
activate HELP from this window.

QUAD-WISE WINDOWS

You can graphically show four dimensional representations with the
Quad-Wise windows.

To activate a Quad-Wise window:

1.   Press the Ctrl and Enter keys at the same time to bring the
Directory window to the front.

2.   Place the cursor within the DJI/Year scatterplot in the
Directory window, and press the pound # key.

The cursor changes to a question mark indicating that the system is
waiting for you to make another selection.

3.   Place the cursor within the Automobile/Year scatterplot (where
Automobile on the left column intercepts with Year on the bottom
row), and press the pound # key to select another pair of
variables.

This action produces two scatterplots with linking lines that
connect all corresponding data points. Note that the low levels of
the DJI have little correlation with the wide range of automobile
sales.

To display subsets of this plot:

1.   Place the cursor within the black space of the Quad-Wise
window.

2.   Press the r key to view the red colored cases.

3.   Press the n key to view the uncolored cases.

4.   Press the a key to view all of the cases.

5.   To close this window, place the cursor within the Quad-Wise
window and press the Esc key. For more information on this window,
activate HELP from this window.

PARALLEL COORDINATE WINDOWS

The Parallel Coordinate window is a simultaneous presentation of
numerous variables. This graphic is handy when examining trends
across variables, longitudinal studies or  repeated measures
designs. You can refine the presentation in several ways.

To access the Parallel Coordinate window:

1.   Place the cursor within the black space of the Directory
window and press the at @ key. The system displays the Parallel
Coordinate window.

2.   Maximize or enlarge the Parallel Coordinate window.

To prune the axes:

1.   Place the cursor within the vertical Automobile axis (bar) and
press the down arrow key.

The Automobile axis (bar) disappears from the plot making more room
for the other axes.

To display subsets of this plot:

1.   Press the r key to view the red colored cases.

2.   Press the n key to view the uncolored cases.

3.   Press the a key to view all of the cases.

4.   Place the cursor within the Parallel Coordinate window and
press the Esc key to close the window.

When you want more information on this window, click on HELP from
this window.

FRACTAL FOAM WINDOW

The Fractal Foam window is a new, graphical display of highly
correlated variables. It is a very quick and easy way to select a
variable from your dataset and see which other variables are highly
correlated with it. Also, relative properties of the distributions
for the variables are graphically displayed:  skewness, kurtosis
and standard deviation. A quick look at the Fractal Foam window for
variables that you intend to analyze can reveal a lot of potential
problems, such as multicollinearity, violations of assumptions of
normality, or data with a correlation structure that doesn't make
sense for a particular model.

1.   Remove all colors by pressing the Home key over the
BMDP/DIAMOND logo at the bottom left corner of the directory
window.

2.   Position the cursor over any Term histogram and press the
exclamation point ! key to set a large rainbow of colors.

Opening a Fractal Foam window using Economy:

1.   Position the cursor over any Economy histogram and press the
period . key.

The system displays the Fractal Foam window for Economy.

The Fractal Foam window (labeled Foam) consists of many bubbles,
each one representing a variable from your dataset. The central,
large, white bubble represents the variable Economy, the variable
you selected in the Directory window. This variable is the focus
variable. The bubbles are color coded. Other white bubbles in this
window also represent Economy.

There are five medium-sized bubbles around the periphery of the
Economy bubble. These bubbles make up the fan-out of the focus
variable and represent the five variables most highly correlated
with Economy. You can change the fan-out number to increase or
decrease the number of bubbles surrounding the focus variable.

The most highly correlated variable always appears at the three
o'clock position, and in this example the most highly correlated
variable is the red GNP bubble. The other highly correlated
variables display counter-clockwise from GNP in descending order of
correlation coefficient. In this example they are: Year,
Population, Term, and DJI.

Each correlated variable has its own fan-out of variables
correlated with it. Those variables also appear in order of
descending correlation, and their number is the same as the fan-out
of the focus variable. You can specify from 0-9 fan-out variables.

To change the fan-out to show only two correlated variables:

1.   Position the cursor anywhere within the Fractal Foam window,
and press the 2 key.

The system displays the two correlated variables, GNP at 3:00 in
red and Year at 9:00 in yellow.

Notice that each of these variables is surrounded by their two most
highly correlated variables. For GNP these are Year, in yellow, and
Population in green.

The size of the bubble for correlated variables is directly
proportional to the absolute value of the correlation coefficient.
The bubble for a variable which has a correlation coefficient of 1
with the focus bubble will be drawn one-half the size of the focus
variable.

Notice that the Economy bubble is elliptical and has its title
running along the major axis slightly uphill as read left to right.
This indicates relative positive skewness. Conversely, negative
skewness is indicated by negative slope. The elliptical effect is
a display of kurtosis, the more relatively kurtotic, the more
oblong the ellipse.

Notice that the GNP bubble has a jagged edge. This indicates that
GNP displays a large standard deviation relative to the other
variables in the dataset.

2.   Reset the display to a fan-out of 5 by pressing the 5 key with
the cursor inside the Foam window.

Notice the number 98 in the bottom left corner of the Foam window.
It is white, indicating that the set of data points is all colors,
and that there are 98 cases present. But you can look at any
colored subset of points.

3.   Press the c (lower case) key, making sure the cursor is in the
Foam window.

This action displays the cyan group for the variable term.

The system redraws the contents of the Foam window with a cyan 16
in the lower left corner, indicating that a Foam window for the
cyan subset, with 16 cases, is being displayed.

Notice that most of the bubbles are smaller than they were with the
full dataset, indicating that correlations are not as great. Notice
also that this is a much different set of highly correlated
variables: Automobile, DJI, Capacity, Unemployment, and Year.

There are other options for changing aspects of the display,
including a different way to draw the sizes of the bubbles, known
as the scaled representation, under HELP.

4.   With the cursor in the Directory window, press the Esc key
twice.

The system displays the BMDP/DIAMOND Startup window.

You have now completed the Fractal Foam window section. At this
point, you can continue on to the next section, or you may quit.

To Quit:

1.   Place the cursor in the BMDP/DIAMOND Startup window, and press
the Esc key twice.

RE-INVOKING BMDP/DIAMOND

You can re-invoke BMDP/DIAMOND on any number of subsets of the
data.

To re-invoke BMDP/DIAMOND on the red colored cases of the data:

1.   Place the cursor within the DJI/Year scatterplot in the
Directory window and press the Enter key.

The system displays the Pair-Wise window for DJI/Year.

2.   Place the cursor on the red number (under Cases or
correlations) in the Pair-Wise window, and press the Enter key.

A new Directory window appears for quarters when DJI was low. Note
the change in the title (Directory for Econ.1 from Directory For
Econ), and the change in the number of Cases. All of the options in
BMDP/DIAMOND are available to explore this new subset of data.

You can save any subset of data by returning to the operating
system and copying the files that were created (see your system
documentation for more information).

3.   Place the cursor within the new Econ.1 Directory window, and
press the Esc key two times to close the window.

EXITING BMDP/DIAMOND

1.   To close the Directory window, place the cursor within the
Directory window, and press the Esc key two times.

The BMDP/DIAMOND Startup window appears again.

2.   To exit BMDP/DIAMOND, place the cursor within the Startup
window, and press the Esc key two times.

SUMMARY

You have now worked with most of the BMDP/DIAMOND windows and
features. You have learned how to color subsets of data points,
show best-fit straight lines, transform variables, manipulate
graphic representations of data, and how to re-invoke BMDP/DIAMOND.

ICE

BMDP/DIAMOND includes Ice, a multi-dimensional visualization tool.
Ice lets you map and display up to nine different variables from
your dataset onto its various attributes. In the following example
you will invoke Ice from BMDP/DIAMOND, map and unmap variables to
different display attributes, manipulate the display, and learn how
to look at interactions between several variables at a time.

To access Ice:

1.   From the BMDP/DIAMOND Startup window, click on econ from the
list of sample dataset names.

The system displays the Directory window for the dataset econ.

To invoke Ice from the Directory window:

1.   Position the cursor in the lower left corner on 98 cases and
press the F2 key.

The system displays the Ice window for the econ dataset, labeled,
Visualization of econ.x.

2.   Maximize or enlarge the new window.

The Ice window contains a large cube marked off with a grid. Inside
the cube are small colored rectangular solids of varying
thicknesses with different sized holes running through them. These
are called glyphs. Both the position of the glyphs within the cube
and the properties of the glyphs themselves display information
about the variables in your dataset.

First, let's look at the large cube. It has an x, a y, and a z
axis. Each axis is partitioned into fifths by the grid appearing on
the back walls.

Imagine that each grid line represents the intersection of a plane
with the wall, such that the cube is divided by five parallel
planes per axis, dividing the cube into 5 x 5 x 5 or 125 little
cells. Variables are mapped to the x, y, and z axes. Each cell then
represents a range in x, y, and z. If there are any cases with x,
y, and z  in the cell's range, Ice draws a glyph when the threshold
is exceeded. (For more information on threshold, see the discussion
of threshold on the next page.)

In the above example, Year, Population, and GNP are mapped to the
x, y, and z  axes of the cube, respectively. The white glyph near
the origin of the cube tells us that there are cases in the dataset
with low values of Year, Population, and GNP. Likewise, the red
glyph in the upper right-hand corner of the cube tells us there are
cases with high values of Year, Population, and GNP. The other
glyphs lie on a line between the low x, low y , low z cell and the
high x, high y, high z cell. The alignment indicates that these
three variables are highly correlated.

To increase the number of cells:

You can divide the cube into a larger or smaller number of cells.
There is a slider bar along the lower edge of the window labeled,
Cells = 5.

1.   Position the cursor along the cells slider bar just above the
number 12 (16 for UNIX users) , and click the left mouse button.

You should now see that Cells = 12 (16 for UNIX users) and that the
display has been re-drawn with 12 (16 for UNIX users) cells per
side instead of five. The cases still appear to occur in a more or
less monotonically increasing fashion in x, y, and z.

To rotate the cube:

1.   Rotate the cube by pressing the up and down arrow keys, the
left and right arrow keys, or the less than < and greater than >
keys, until you have seen the display from all angles.

NOTE:     If your display can show enough colors, you may notice
that the back sides of the glyphs are darker than the front sides.
This can make it easier to see their orientation.

2.    Press the Home key to restore the cube to its original
position.

Using the other slider bars

There are two other slider bars, one for Population and one for
Variation. The population slider bar allows you to set a threshold
for drawing a glyph in a cell. It is currently set at one, which
means Ice will draw a glyph in a cell if there is more than one
case in the cell.

1.   In preparation for setting the Population slider bar, adjust
the Cell slider bar until it shows Cells=5.

To increase the Population threshold:

1.   Position the cursor in the middle of the Population slider bar
and click the left mouse button.

The system sets the Population threshold to about 10 and now you
see four glyphs. The other slider bar, Variation, is used to set a
threshold for the variation within the cell.

The process of partitioning the data into cells requires both of
these new measures, population and variation. By default, Ice
displays information about the population and the variation for
each cell. Population is depicted as glyph thickness. The white
glyph in the lower left hand corner is a thick one (a highly
populated cell), while the magenta glyph next to it is somewhat
thinner (lower population). Variation is depicted by glyph
fullness; with high variation, the glyph is full, and the size of
the hole running through the glyph is small.

The magenta glyph is a full one (large variation) while the white
glyph rates lower in fullness (lower variation). You can remap
variation and population to other display attributes, or unmap
them, freeing the two attributes to display other variables from
your dataset.

There are two lists in the upper right corner of the window. The
one on the left is the list of attributes for the Ice display. The
list on the right contains variables from the dataset, plus the two
new measures (Variation) and (Population). When you map a variable
to a display attribute, the variable name appears across from the
attribute's name. For example, Year is mapped to the display
attribute x, the x dimension of the cube. The last attribute on the
list is hue. The value of hue increases from white to red, as shown
by the progression of colors next to the name hue. The variable
names that appear below hue and its mapped variable, Economy, are
the unused variables.

To unmap the Variation variable from the display:

1.   Position the cursor over the variable Variation in the upper
right corner of the window, and press the space bar.

This action sends the Variation variable to the bottom of the
unused variables list.

To map a different variable to fullness:

1.   Position the cursor over the variable Unemployment, and press
the key corresponding to the first letter of the attribute's name,
in this example, f for fullness.

Notice that the highest levels of unemployment (greatest
fullness/smallest hole) occur in the blue colored cell, and the
lowest levels (largest hole) occur in the white and red cells.

To replace current mappings:

This example shows how to replace (Population) with Prime for the
attribute thickness.

1.   Position the cursor on the variable name Prime and press the
t (for thickness) key.

This action maps Prime to thickness and sends (Population) to the
unused variables list.

Notice that the highest level of the Prime interest rate also
occurs in the blue colored glyph with lower values in the other
glyphs.

There are three more display attributes. These all have to do with
the position of the glyphs in their cells. They are glyph azimuth,
elevation, and roll.

To map a variable to sideways rotation:

1.   Position the cursor over DJI and press the a (lower case) key,
for azimuth.

Notice how the glyphs have rotated within their cells. The white
glyph shows the least rotation, having the lowest values of the
DJI. The red glyph shows the most rotation, having gone through
more than 180 degrees so that we now see its back side which may be
indicated by the darker, duller shade of red.

To toggle the mapping of azimuth on and off without remapping it:

A very useful feature allows you to toggle the display of any
attribute on and off.  This allows you to really see where the
mapping has large and small effects. To see more clearly the
largest and smallest values of its associated variable, DJI, you
can toggle the mapping of azimuth on and off.

1.   Position the cursor over the attribute name azimuth and
repeatedly press the slash  / key.

Look how far the red glyph turns. The red glyph turns so far that
you see its back side. Notice that when the attribute is toggled
off, the variable name is grayed out. Leave the attribute azimuth
toggled off.

To map a variable to tilt upwards:

1.   Position the cursor over the variable Automobile and press the
e key (for elevation).

2.   Position the cursor over the attribute elevation and toggle it
off and on by pressing the slash / key to see the change in
elevation.

3.   Leave the attribute elevation toggled off.

To map a variable to roll:

1.   Map the variable BondShort to the attribute roll by
positioning the cursor over the variable BondShort and pressing the
r key (for roll).

2.   Toggle the mapping of roll on and off with the slash / key to
see the change in roll.

3.   Leave the attribute roll toggled on.

To map three variables to orientation:

1.   Toggle the mappings for azimuth (DJI), and elevation
(Automobile) back on.

There are now nine variables simultaneously mapped to attributes of
the Ice display.

By rotating the cube and toggling attributes off and on, you can
get a clear picture of the relationships between many variables in
your dataset. Ice is a very powerful tool for simultaneously seeing
these relationships for more than the customary two or three
variables. You should spend some time working with Ice on a dataset
that you know well so that you can really learn how to fully use
this powerful visualization tool.

2.   With the cursor in the Ice window, press the Esc key twice.

3.   With the cursor in the Directory window, press the Esc key
twice.

The system displays the BMDP/DIAMOND Startup window.

You have now completed the Ice section. At this point, you can go
back to any of the previous sections, or you may quit.

TO QUIT

1.   Place the cursor in the BMDP/DIAMOND Startup window, and press
the Esc key twice.