A methodology for structured subjective adjustment of quantitative forecast

This research surveys the methods presented in the literature by which subjective information is integrated with quantitative forecast models. One of the methods, the subjective adjustment of quantitative forecasts, is examined for areas of potential improvement. As currently practiced, subjective adjustment suffers from a lack of procedural structure and a lack of reproducibility. This research presents a methodology developed to address specifically these primary shortcomings of subjective adjustment.;The proposed methodology provides a structured process by which the forecaster introduces contextual information into a quantitative forecast model. The methodology predicts the effects of intervening events on a base quantitative model by observing the effects of previous, contextually similar events. Intervening events are grouped into context classes by the forecaster. The aggregate (average) effects of each context class are used to predict the effects of anticipated intervening events of similar contextual nature.;Validation of the proposed methodology is addressed through a series of statistical analyses of the application of the methodology to sets of actual time series. The statistical analyses are performed using two implementations (Phase I and Phase II) of the methodology and assess the significance of the improvement in accuracy due to the application of the methodology. The Phase I implementation of the methodology considers only the contextual categorization of intervening events, whereas the Phase II implementation also allows for the subjective characterization of intervening events. Three analyses are performed with the Phase I implementation of the methodology using the author as the forecaster. One analysis is performed (in three separate studies) with the Phase I implementation using volunteer forecasters. The analyses demonstrate improvements in accuracy of modest significance. The Phase II analyses further illustrate the superiority of aggregate forecasts over their constituent forecasts, a finding evidenced in the literature.;The applicability of the proposed methodology to industry is addressed by means of interviews with four experienced industrial forecasters. The industrial forecasters examined the proposed methodology in terms of both structural logic and applicability to their particular forecasting environments. All of the forecasters interviewed agreed with the structure and logic of the methodology and three of the four forecasters expressed belief that the methodology would help them to meet their forecasting needs.