Graphics Reference
In-Depth Information
of the variables should be displayed. his allows extra flexibility in terms of the
choice of derivatives to display.
Filtering curves: Users can filter the curves to see only auctions of interest byusing
filter widgets called TimeBoxes. One can click on the TimeBox icon of the tool-
bar and draw a box on the time series panel of interest. Every curve that passes
through the box (between the bottom and top edges of the box for the duration
that the box occupies) is kept, while all of the other curves are filtered bygraying
them out. hecorresponding auctions are also removed fromthe attribute panel
on the right. Figure . shows a typical filter TimeBox used to view only the
auctions that end with high price velocity. In the attributes panel, users can see
that all of these auctions finished around the weekend. hey can apply multiple
TimeBoxestothesameorseparatevariablesinordertoformconjunctive queries
(i.e., a logical AND combination of individual TimeBox queries). For example,
users can search for auctions that end with low prices and high velocities.
Searching for patterns in curves: When comparing price curves and (especially)
price dynamics, one useful tool is the pattern search. his is achieved by draw-
ing a SearchBox on a selected curve for a certain time duration. he pattern is
the part of the series that the SearchBox covers horizontally, and this SearchBox
is searched Simultaneously for all other curves at any time point in the auction.
here is a tolerance handle on the right of the SearchBox that allows the simi-
larity to be specified. For example, users can search for auctions that have price
curves with steep escalations at any time during the auction. TimeBoxes and
SearchBoxes can be combined into multistep interactive searches.
Functional summaries: One can obtain numerical summaries for a set of func-
tional objectsusingthe riverplotinFig. . .heriverplot isacontinuous formof
theboxplotanddisplaysthe(pointwise) median togetherwith the %and %
confidence bounds. he riverplot allows for a condensed display of the average
behavior of all curves together with the uncertainty around this average.
Attribute Operations
Manipulating the attribute data and observing the coupled functional data are useful
ways of learning about relationships within the data across the different data types.
he following operations support such explorations (in addition to more standard
explorations of attribute data alone).
Sorting auctions: Users can sort the auctions by any attribute by clicking on the
attribute name in the first row. A click sorts in ascending order, while the next
click sorts in descending order. Sorting can be performed on numerical as well
as text attributes. hesorting also recognizes day-of-the-week and time formats.
he sorting is useful for learning about the ranges of the values for the different
attributes, the existence of outliers, the absence of certain values, and possible
errorsandduplications inthedata. Furthermore,sorting can allowuserstovisu-
ally spot patterns of “similar" auctions, bymaking auctions with similar attribute
values appear consecutively in the auction list. Users can sort according to more
Search WWH ::




Custom Search