Important deadlines:

Submission (closed)
May 11, 2014
Acceptance notification
June 15, 2014
Payment of fee
July 10, 2014

Last modified:
December 27, 2021 15:49:06 (gk)

Mike West: Forecasting, Portfolio Decisions, Model Uncertainty and Combination – Some Historical Perspectives and Some Current Challenges (Keynote Talk B, Thursday, 17:00)

One of the founding fathers of modern Bayesian analysis, Dennis Lindley (1923-2013), was a lifelong advocate of the integration of decision analysis and inferential/modelling studies. From the early 1990s, Lindley repeatedly expressed his view that Bayesian decision theory was increasing under-represented in modern R&D in the Bayesian community. I believe that his comments (from the early-1990s) are no less appropriate, and relevant, today, and that a renewed focus on decision theoretical-perspectives can aid in advancing applied Bayesian work in various fields.
This talk will be a conversation touching on a number of general questions linked to decision analysis perspectives, drawing on my own work in various contexts. Topics will be (selectively) drawn from: (a) current, novel developments in Bayesian decision theory emerging from personal and corporate interests in improved financial portfolio construction that involve multi-step forecasts and decisions in financial time series; (b) a broader view of Bayesian model evaluation and comparison when multi-step forecasts underlie primary decisions, and issues arising with purely statistical Bayesian model averaging that does not include decision contexts; (c) dealing with multiple candidate models and coherent Bayesian model combination in a forecasting context; and (d) – if time and interest permit – related questions of aggregation of forecasts in "group forecast summary/decision making" contexts. All topics relate to challenging, current research questions that tie back to relevant published work as old as 30 years now, so some historical perspective will be ingredient.