A1041
Title: Probabilistic forecast aggregation with statistical depth
Authors: James Taylor - University of Oxford (United Kingdom) [presenting]
Abstract: Interval and distributional forecast aggregation methods are considered that can be applied when there are many forecasters, and their past accuracy is unavailable. The median and trimmed mean have been proposed as robust alternatives to the mean. For interval forecast aggregation, the median and trimming methods consider each bound separately. To try to use the available information better, the bounds are treated as bivariate points with statistical depth used to order the points in terms of centrality. The deepest point can be viewed as the median interval forecast, and the depth of each point can be used as the basis for trimming. For distributional forecasts, the literature presents aggregation methods for which the median or trimmed mean is obtained separately at each point of the support of the distribution. However, if one part of a distributional forecast is outlying, the appeal of using the rest of it is perhaps reduced. Functional depth is used to provide a measure of centrality for each distributional forecast, and hence identify the deepest function, which can be viewed as the median forecast. Functional depth is also used as the basis for trimming. An empirical illustration is provided using data from surveys of professional macroeconomic forecasters.