Graphics Reference
In-Depth Information
; Shmueli et al., ). In Sect. . we describe an interactive visualization tool
designed for the display and exploration of functional data. We illustrate its features
and benefits using the example of price curves, which capture the price evolution in
online auctions.
he insightful display of functional data comes with many, many different chal-
lenges, and we are only scraping the tip of the iceberg in this essay. Functional data is
challenging due to its high object dimensionality, complex functional relationships
andthe concurrency present among the functional objects. Wediscusssome ofthese
extra challenges in Sect. . .
Online Auction Data from eBay
5.2
eBay (www.eBay.com) is one of the major online marketplaces and currently the
biggest consumer-to-consumer online auction site. eBay offers a vast amount of rich
bidding data. Besides the time and amount of each bid placed, eBay also records
plenty of information about the bidders, the seller, and the product being auctioned.
Onany given day,several million auctions take placeon eBay,andall closed auctions
fromthe last days are made publicly available on eBay's website. his huge amount
of information can be quite overwhelming and confusing to the user (i.e., either the
seller,the potential buyer, orthe auction house) that wishes toincorporatethis infor-
mation into his/her decision-making process. Data visualization can helptoalleviate
this confusion.
Onlineauctionslendthemselvesnaturallytotheuseoffunctionaldataforavariety
ofreasons. Online auctions can beconceptualized asaseries of bidsplaced overtime.
hefinite timehorizon of theauction allows the study ofthe priceevolution between
the start and the end of the auction. By “price evolution” we mean the changes in
the price due to new bids as the auction approaches its end. Conceptualizing the
price evolution as a continuous price curve allows the researcher to investigate price
dynamics via the price curve's first and second derivatives.
It is worth noting that empirical research into online auctions has largely ignored
the temporal dimension of the bidding data, and has instead only considered a con-
densed snapshot of the auction. hat is, most research has only considered the end
of the auction by, for example, concentrating only on the final price rather than on
the entire price curve, or by only looking at the total number of bidders rather than
the function describing the bidder arrival process. Considering only the end of the
auction results in information loss, since such an approach entirely ignores the way
in which that end-point was reached. Functional data analysis is a natural solution
that allows us to avoid this information loss. In a recent series of papers, the first two
authors have taken afunctional approach and shown that pairing the priceevolution
with its dynamics leads to a better understanding of different auction profiles (Jank
and Shmueli, ) or to more accurate forecasts of the final auction price (Wang
et al., ).
Search WWH ::




Custom Search