No matter what the situation, getting value from real-world data almost always involves the use of software. Any two pieces of software are necessarily different, so the often-heard question “how does software A compare with software B?” can be tricky to address. Probably the only way to make progress towards an answer is to assess the relative utility of A and B in the context of a specific challenge, and in the hands of a specific user, and this realisation is part of the motivation for the ENBIS challenge, which will be issued annually. The other motivation for the challenge is (over time) to gain insights into which textbook approaches to analysis actually prove most useful in the context of real-world data arising from some specifically chosen context or problem domain.
Each year, ENBIS will issue a challenge a few months prior to our autumn conference. Anyone can offer a submission, and details on how to do this are below. All submissions will be reviewed by a panel of three adjudicators from ENBIS, who will jointly decide on which is the best overall.
Firstly, because it will be informative and fun. Secondly, (assuming the winner is domiciled in Europe) ENBIS will pay your travel costs, hotel costs and the fee to attend our autumn conference. The winner will also have a fifteen-minute slot in the conference agenda within which they can present their winning submission.
The 2011 ENBIS Challenge by JMP was entitled »Maximising Click Through Rates on Banner Adverts: Predictive Modeling in the On Line World«. The challenge consisted of attempting to increase advertising revenues to the portal by dynamically surfacing specific advertisements that particular types of people would find attractive and were more likely to click on. There was a body of observational data that described both the profile of a user and what free content they had viewed in the past, so the hope was that a useful model could be built that will predict the profile of a user from their viewing habits alone. If yes, then appropriate, personalised adverts could be displayed.
Those who undertook the challenge played the role of a statistical advisor that has been contracted by Content4U to undertake the required predictive modelling using the software of their choice. As is often the case, the consulting brief was a little loose, and, to make matters worse, the database admin who provided the data had just gone on holiday, and was not planning to return until after the deadline for the final presentation. So some informed guesses in terms of what exactly was the client trying to achieve were necessary in the process.
The winner of this Challenge was announced at ENBIS-12 in Ljubljana (Slovenia): Rainer Kempkes.
The 2010 ENBIS Challenge by JMP was about optimization, more specifically about maximizing profit when resources are limited.
Using a simulation of a complex, real-world situation (an LPCVD deposition step from semiconductor manufacturing), the challenge consisted of finding the best process recipe for three responses and nine factors. The hierarchical product structure and the way the responses are measured surfaced some interesting issues of data exploration and the design and analysis of an experiment. But the bigger challenge arose from the choice of a sequential experimental strategy that would allow for rapid learning and still leave some resources left over to run production and actually make money.
The simulator required JMP to run, but software other than JMP could also be used for analysis and design.
The winner of this Challenge was announced at ENBIS-11 in Coimbra (Portugal): Marion Chatfield of GlaxoSmithKline Services Unlimited.
Maximising Click Through Rates on Banner Adverts: Predictive Modeling in the On Line World. Find out more…
The 2010 ENBIS challenge was about optimizing an industrial process via experimental design using a simulator to get realistic process results from a chip-wafer process. Find out more…
Developing Products and Services People Want To Buy: Visual Analytics for the Analysis of Marketing-Related Data; Accell Inc.” . Find out more…