Information Technology Reference
In-Depth Information
items between the two in the plot - for example, in the right image of Fig
2.8, GE is on the opposite side of the mass of links from WMT and APPL,
but all three are highly correlated. While interaction can partially address
this issue, there could still be a perceptual bias for items at the centre
versus items at the edge.
Fig. 2.8.
Left: 100 stocks showing 10,000 correlations in a 100x100 grid. Right,
same ten stocks as Fig. 2.7 right, visualized as a graph where proximity represents
the strength of the correlation, e.g. GOOG and IBM (near top left) have a high
degree of correlation but GOOG and PG have an inverse correlation.
Correlations depicted as a force-directed scatterplot on a sphere instead
of a plane have potential benefits:
•
No items will be at the edge of the plot. Perceptually, no item can
be located in the centre or a perceptually preferred position over
any other item.
•
Inverse correlations on a force-directed sphere-based mapping will
tend to be as far apart as possible, which intuitively is the opposite
side of the sphere. This could be useful in forming an intuitive
mental model of the correlation space, with strong correlations
close-by, inverses most likely to be on the opposite side, and weak
correlations likely to be orthogonal.
Based on the above reasoning, a few design sketches of the
visualization were created outlining the required features (e.g. layout,
markers and navigation) and desired features (e.g. filtering and narrative).
The visualization was then implemented using an iterative approach,
adding incremental features as development advanced.
Search WWH ::
Custom Search