Simple robust averages of forecasts: Some empirical results

By Victor Richmond R. Jose, Robert L. Winkler. 2008

241 citations recorded by the source corpus.

Subfields: Economics and Econometrics, Global and Planetary Change, Management Science and Operations Research

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Economics and Econometrics

In Global and Planetary Change

In Management Science and Operations Research

Citations in the corpus

Combining Forecasts: An Application to Elections (2014, unknown): “Jose and Winkler (2008) provided similar results for combinations of five, seven, and nine forecasts.”
Toward Automatic Time-Series Forecasting Using Neural Networks (2012, methodology): “GRNNs, that is, three and five, where the five spread factors for five GRNNs are 5th, 25th, 50th, 75th, and 95th percentiles, respectively, of the nearest distances of all training samples to the rest of samples; and three levels for fusion methods, that is, mean, median, and trimmed mean [54].”
Neural network ensemble operators for time series forecasting (2014, background): “Jose and Winkler (2008) using a similar argument advocated the use of trimmed and winsorised means.”
Meta-learning for time series forecasting and forecast combination (2010, methodology): “This is in accordance with guidelines given in [24], where 10-30% are recommended.”
Forecasting: theory and practice (2020, unknown): “ical studies (Elliott et al., 2005; Jose and Winkler, 2008; Andrawis et al., 2011; Kourentzes et al., 2019).”
'Horses for Courses' in demand forecasting (2014, methodology): “Many recent studies have verified that the combination of methods leads to more accurate forecasts, whilst, at the same time proposing more sophisticated weightings such as the trimmed and Winsorized means (Jose and Winkler, 2008), and the use of information criteria (Taylor, 2008; Kolassa, 2011).”
Improving forecasting by estimating time series structural components across multiple frequencies (2014, methodology): “…weight selection processes for combining forecasts (He and Xu, 2005; Taylor, 2008; Kolassa, 2011); however, simple approaches are considered to be robust and perform reasonably well relative to more complex methods (Clemen, 1989; Hibon and Evgeniou, 2005; Timmermann, 2006; Jose and Winkler, 2008).”
A comparison of AdaBoost algorithms for time series forecast combination (2016, background): “The empirical performance of forecast combination has been substantiated in various experiments (see e.g. Aksu and Gunter 1992; Gunter 1992; Stock and Watson 2004; Clements and Hendry 2007; Jose and Winkler 2008), in prominent empirical case studies in operational research (see e.”
The art of modelling range‐shifting species (2010, background): “Other research fields use ensembles, often by taking averages over all available models that satisfy certain prespecified criteria – e.g. tests of predictive skill, or of model independence (Tebaldi & Knutti 2007; Jose & Winkler 2008; Abramowitz 2010).”