Up to this point all statistical summary information has been abstractly represented as nodes, or regions, with a gray scale. We never got to see the gray scale translated into exact arithmetical data (percentages) or into a concrete population (observations). This aspect has been postponed until the current section.
The diagrams in Abundantia Verborum are more than passive pictures. They can act as gateways to specific subsets of the data. Nodes, or regions, are hotspots to their population. Clicking on such a node or region, which is called zooming in on that diagram part, calls up the Observation Browser, containing only the population of that node or region. Apart from the list of these observations the Observation Browser also signals, in percentages, how big its current contents is with respect to the complete filtered workshop and relative to the complete unfiltered workshop.
In "demowork.wrk" create a Venn diagram with a threshold at zero percent, and with the following displayed labels:
SEM:having old ageSEM:not most recent typeSEM:no longer existingnot(1), not(2)
and not(3) collectively mean.
The numbers 1, 2 and 3 refer to the
indices in the "Displayed labels" panel. If you do not know these
by heart, you can always move the Observation Browser, so that
the "Displayed labels" panel becomes visible (you click on the title
bar of the dialog box, and while keeping the left mouse button down you
move the mouse; finally you release the mouse button).
Clicking anywhere inside the ellipses results in the behaviour
you intuitively expect. Experiment and pay special attention to the
zoom information panel in the Observation Browser!
Now click on the "Graph" panel with your right mouse button and select
the diagram type "Inverted Hasse diagram". This diagram type is exactly
the same as the default Hasse diagram, except that it is upside down.
This time {} is on top. The reason for introducing a separate
diagram type for such a trivial difference is that in different disciplines
different orientations are used, and that, according to our experience,
for such matters people tend to have a strong preference for what they
are already familiar with.
In Hasse diagrams, as opposed to Venn diagrams, the background is not an integral part of the actual diagram. Clicking on the background is equivalent to "Workshop | Browse Observations...". Further all the nodes are hotspots to their population. Once again we invite you to experiment and, once again, pay attention to the zoom information panel in the Observation Browser.
Once again we invite you to click on the "Graph" panel with your right
mouse button, this time to select the diagram type "Schematic Diagram (Vertical)".
This diagram type is exactly
the same as the horizontal diagram, except that it has rotated 90 degrees.
This time root is on top. The reason for introducing a separate
diagram type for such a trivial difference is similar to what we just
said in the context of Hasse diagrams.
In Schematic diagrams, as in Hasse diagrams, the background is not an
integral part of the actual diagram. Clicking on the background is
equivalent to "Workshop | Browse Observations...".
And so is clicking on the "root", which as we stated
in the last lines of
3.3.2 Venn, Hasse and Schematic diagrams,
represents the complete (filtered) workshop.
Further all the nodes are hotspots to their population.
Click on "1"! In the Observation Browser you notice that
the formulations in the zoom information panel are a little different this
time. Now the information reads: 1,
either 2 or not and
either 3 or not. This is indeed in accordance with
how we defined the population of Schematic diagram nodes in
3.3.2 Venn, Hasse and Schematic diagrams.
As with the other diagram types, we again invite you to experiment.
In 3.2.4 Adding labels via zooming we anticipated zooming and mentioning that zooming, like filters, can be used for data classification with "Tag All". In that section we referred to the current section, saying that here we will thorough explain zooming, explain how it related to filters and explain how it can be used to refine data classification. What zooming is has been the topic of the previous parts of this section. How it relates to filters, can be inferred from 3.3.5 Filtered diagrams because that section explains how graphs and filters can be combined, and zooming is just an extra dimension of graphs. How, finally, can zooming be used to refine data classification ?
The answer is: the same way as
filters can, with all additional benefits from combining filters with
graphs (cf. 3.3.5 Filtered graphs).
Refining data classification can mean two things. First it
can mean cleaning up the classification and making it more uniform. In
3.2.5 Manually adding labels we mentioned that
it typically takes several passes to obtain a stable, satisfactory
set of labels in groups such as SEM or SAID-OF.
The process of cleaning up and uniformisation can be steered by filters
and graphs. By filtering, zooming or both you can collect all relevant observations
in the Observation Browser, and after comparing the
examples, you can perform cleaning up and uniformisation through
"Tag All" and by removing obsolete labels with the
Label Browser.
Secondly, refining data classification can mean adding new labels
on the basis of formal criteria, in order to simplify subsequent
analyses. Suppose for instance the observations matched
by some complex condition such as AND(NOT(X),OR(Y,CONTENTS(Z)))
(X and Y being some labels
and Z being some query)
are a special group that turn up over and over again in your analyses,
then it might be interesting to create a label called
e.g. <analysis aids>:<special group so and so>.
By filtering, zooming or both you can collect all relevant observations
in the Observation Browser and with "Tag All" you can give them the label
<analysis>:<special group so and so>.
In subsequent analyses you can then use this one label
instead of the complex specification.