Canberra Evaluation Forum Archive
Rosemary A Bailey, Professor of Statistics, Queen Mary University of London, will give a seminar on " Conflicts between optimality criteria for block designs with low replication" , 25 November 2012 at 2.30pm.
Peter Harper, Deputy Australian Statistician delivered a presentation to the Canberra Evaluation Forum on 15 October 2009 on the importance of statistics to evidence based policy and in particular, the use of appropriate statistics and how they can sometimes be misused.
His presentation also covered the role of the Australian Bureau of Statistics (ABS) as Australia's national statistical agency in providing policy makers with high quality statistics and outlined how the ABS works in partnership with others to enhance the information base for Australian decision makers. Putting the Evidence into Evidence Based Policy | PowerPoint Presentation (1.1MB)
Estimation of Customer Lifetime Value in Service Contracts
Professor Andreas Ruckstuhl, Zurich University of Applied Sciences, visiting Alan Welsh in Statistical Science gave a seminar on “Estimation of Customer Lifetime Value in Service Contracts” on Thursday November 18 2010.
Abstract: Customer analytics is a process by which data from customer behaviour is used to help make key business decisions in direct marketing, customer relationship management, site selection, etc. An important concept in such kind of analytics is the customer lifetime value (CLV) where individual customers are valuated with respect to their future expected profits.
We introduce an approach in which such a CLV calculation (actually estimation) can be implemented for a complex contractual setting as in, e.g., newspaper or telecom subscriptions. The approach is based on a Semi-Markov formulation of customer dynamics. As a consequence, the sojourn time of customers in pre-specified states and the transition probabilities must be estimated and predicted based on the available information in the business data warehouse. The estimation is done using a proportional hazard approach for the sojourn time and multinomial regression model for the transition probabilities. As such prediction must be repeated regularly, one strives for an (semi-) automatic analysis. To do so, we introduce the idea of a task and of “task cards” which keep track of the meta-parameters of different tasks
Return to Events Archive