been developed for analysing market share changes
and monitoring portfolio performance, all using interactive
graphics. The focus lies on changes between time periods
using skyline plots, a new graphic which shows % changes,
absolute changes and shares in one coherent plot.
As an example consider
looking at the DAX, the German stock market index.
(Data available from the webpage:
). The DAX comprises
the 30 German shares with the biggest stock market values,
including Daimler Chrysler, Deutsche Bank, Siemens and
Volkswagen. It is similar to the Dow-Jones-Index
in New York and the Nikkei Index in Tokyo. The
30 shares may be grouped into eight main sectors: Consumer
& Retail, Insurance, Chemicals & Pharmaceuticals, Banking
& Finance, Automobile & Transport, Utilities & Telecommunications,
Software & Technology, Machinery & Industrial.
Fig. 1 shows the relative change of market value for each of the
eight groups between September and October 2001. (The names
have been left out to reduce clutter.) Apart from
Software & Technology with a gain of 6.7% every
sector lost up to 30% during the period, almost certainly
due to the after-effects of September 11th. The three
worst sectors were the first column, Consumer
& Retail with a loss of 30.0%, the second, Insurance (-27.3%),
and the fifth, Automobile & Transport (-22.3%).
This is not so surprising as these are sectors heavily influenced
by such a catastrophe.
Fig. 1: Diagram with relative Changes in the
The diagram is
misleading in that all groups are treated as if they are
the same size. In fact Consumer & Retail is a
fairly small sector compared to the two others: Automobile
& Transport is seven times bigger, and Insurance
ten times bigger. In spite of this, it is common to compare
only relative changes and not take absolute size
In skyline plots the x-value represents the size of the sector
(so that the widths of the columns are no longer equal) and
the relative change of each sector is plotted on the
y-axis. Relative changes are represented by the heights
of the columns and the absolute changes by their areas
(= relative change * size). Figure 2 shows the skyline
plot corresponding to Figure 1. The order and the heights of the
columns are the same, but the width is proportional to size.
Consumer & Retail in the first column (-30.0%) can now be
more readily compared with the ten times bigger Insurance sector in
the second column (-27.3%). The mean change for the DAX as a
whole is marked by the broken line (-13.4%).
Fig. 2: RENOIR-Diagram takes relative and absolute changes
shows the big difference in
absolute changes between Consumer & Retail
and Insurance or Automobile & Transport.
In spite of a similar relative loss, the change
in the last two sectors influences the total
DAX value much more than Consumer & Retail:
the absolute changes are nine and five times bigger
- in the skyline plot this can be seen from the bigger
It is not the exact size
of the absolute change that is interesting
(areas cannot be compared as easily as heights),
but the area should give you a rough idea of
the importance of each change. The relative change
is a good criterion for identifying successful groups, but in
order to get the most important change you have to
take the absolute change into account. Both variables
Fig. 3: Diagram sorted by absolute
It can be very useful to sort
the columns by their areas (the absolute changes) (menu
'Data' -> 'Sorting' -> 'absolute Changes
within Groups') or by their heights (the relative changes) (menu
'Data' -> 'Sorting' -> 'relative Changes within Groups').
The three sectors discussed above have been highlighted
in red to identify them quickly. The smaller effect of the Consumer
& Retail sector and the much larger losses of the other two
sectors can be seen at once. It is also clear that Insurance
and Automobile & Transport are the two sectors with
the most negative influence on the DAX in this period.
Querying and Selection
and selection is essential for skyline plots because
there can be so many columns. Both default and detailed querying
(mouse move with key 'i' or 'd' pressed) are available. Selecting
a column by a dragged mouse move, it is marked in red.
Fig. 4: Diagram, drilled down in Automobile
Within the DAX the sectors had
different relative changes and within the sectors
so do the various shares. By clicking on a column in the
original diagram you can drill down into the sector. In Figure
4 for Automobile & Transport you get the changes for Volkswagen,
Preussag, Lufthansa, BMW, Deutsche Post and Daimler
Chrysler. The grey 'shadow plot' in the background shows
the size of the parent sector and hence which columns belong
You can drill down (up) in all
sectors by pressing the down (up) arrow key.
Fig. 5: Zoom of Sector Automobile
To get more details you can
zoom in with selection of the sector with the middle mouse key as in
Figure 5. A new window opens with a diagram containing
only the zoomed sector (Figure 6). A bird's eye view
is automatically generated in the floating parameter window
where the red rectangular shows which part is currently zoomed.
Looking at the columns widths
for shares in Automobile & Transport, Daimler Chrysler
(highlighted) seems to be nearly as big as the rest of the
stocks together. With a loss of only 13.9% it performed much
better than the other stocks in the sector. Lufthansa lost
25.9% (third column), Volkswagen 28.4% (first column).
BMW had the worst result of -34.2% (fourth column). Querying
shows, that the absolute loss of Daimler Chrysler is bigger than
the loss of BMW (compare the areas). The relatively 'good' result
of Daimler Chrysler and its large share of the sector raises
the result of the sector as a whole up to -22.3%. Without Daimler
Chrysler the sector would have performed worse than Insurance (-29.2%).
Fig. 6: Diagram with zoomed sector Automobile
& Transport and highlighted Daimler Chrysler
Fig. 7: Parameter window with bird's eye view
Changing Categories and Time Period
The parameter window shows the bird's eye
view, a list of the variables in the data set and the two
time periode compared. The time periods can be changed by the
pop-up option menus. The order of the categories can be changed
by drag and drop.
In Figure 1 the same data are highlighted
as in the zoomed window. This shows Daimler Chrysler
as a proportion of the whole sector: only 28.3% of the
change in Automobile & Transport is due to Daimler
Chrysler, but the company's weight in the sector was nearly
Fig. 8: Linked Diagram with highlighted
Fig. 9: Analysis of Structure Diagram
Analysis of Structure
Raw data rather than shares
and changes can be displayed in an 'Analysis of
Structure' plot. Here the y-height is the size in October
2001, and the stocks are plotted in columns of equal width
along the x-axis, but cumulated into the sectors as before.
The second column, Insurance, is the biggest sector.
The first column Consumer & Retail is one of the smallest
sectors, while Automobile & Transport is somewhere in
the middle. As Daimler Chrysler is still highlighted, its
relative size to its sector Automobile & Transport of 50.1%
can be seen by querying.
Sorting in Total
Drilling down in all sectors at the same
time and sorting them by 'absolute Changes in Total'
shows which stock had the most influence (whether positive
or negative) on the DAX. The columns highlighted here are:
Allianz, Munich Reinsurance (both belong to the Insurance
sector) and Deutsche Bank with losses and Siemens with the
biggest gain. Sorting by relative changes gives us the most/least
successful stocks (relative to their size).
Fig. 10: Diagram drilled down in all sectors
with Sorting Menu
Fig. 11: Diagram sorted by absolute change, the
three worst stocks are highlighted (Allianz, Munich Reinsurance,
Change Over Time
But how is the development over several periods
of time? And how bad is this month compared to
other periods? For these questions, a plot with change
over time is necessary: select the 'Change over time'
box in the parameter window and pull the new category 'Change
over Time' up to the first position as in Figure 12. Then
we get the diagram on the right, showing the changes of the DAX
for each period using alternating backgrounds to seperate them.
The periods start with 11/2000-12/2000, 12/2000-01/2001,...
and end with 10/2001-11/2001. But as the deviation from the
mean during all the periods is very high, we should scale the
axis first, before analysing the diagram.
Fig. 12: Selecting Change over Time in the parameter
Fig. 13: Diagram with Change
Fig. 14: Change over Time Diagram with
Scaling of Axis
By clicking outside the right hand border of the
diagram we can rescale to the height of 94.02% as
a new maximum value. The result is shown in Figure 14. The
second last period is the one we analysed before: September
to October 2001. Now it is obvious, that in this period
the DAX lost a lot, but compared to the total development
and the loss in the other periods it was little: the loss in the
time after September 11th was very bad, but the profit, which
was made after it until the end of the period, 'balanced' it again.
There is some evidence of cyclic behaviour. If there is a big
change in one period, there is a reaction in the next.
Examining Daimler Chrysler over time in Figure 15 by
querying and selection suggests that the variability of the smaller
companies in the sector was greater: the bigger losses they
suffered in the immediate aftermath of September 11th were compensated
by the bigger gains they made at other times.
Fig. 15: Diagram with the development
over time, the stock Daimler Chrysler is red highlighted
If you are interested in experimenting with RENOIR yourself,
click on 'Applet' to open a test-version where the DAX-Dataset
can be analysed by clicking on menu 'File' ->
RENOIR can be used for
analysing other kinds of dataset as well: financial
control, risk management, market share analysis,
and election results. Wherever you have positive
data, changing over time, grouped in different
categories such as regions, sectors, business areas
or what ever, RENOIR can help you.
If there are further questions which cannot
be answered by the help (menu 'Help' -> 'Help - Shortcuts'), contact
(C)Copyright 2003 Dept. of Computer Oriented Statistics
and Data Analysis,
University of Augsburg, All rights reserved