Abstract
Professional oboe players almost always have to make their own reeds, which involves a time-consuming process often fraught with wasted effort and discarded results. About one fourth of the total time spent on a reed involves getting it to a stage where it can be tried out on the oboe. Regression analysis was used to aid in making predictions about the ultimate quality of a finished reed based on data available at the initial try-out. The inputs to the regression model consist of several different characteristics of the cane used in making the reeds, and an assessment of the reed in its early stages through this initial try-out on the oboe. The goal to be able to decide whether or not to continue to work on the reed past this stage, based on the predictions of the regression. Thus far, the outcomes predicted by the regression have coincided reasonably closely with the actual outcomes in trials. Several regression models were tried, ranging from pure linear to curvilinear models that include interaction terms and/or squared terms. A particular curvilinear model was deemed the most appropriate.
