APPLICATION OF STATISTICAL REGRESSION TO YIELD MODELING.

The literature on yield modeling has primarily focused on the proper shape of the percent yield versus DA curve (defect density times area curve). One of the conclusions which can be safely drawn is that more than one of these curves can accurately describe data in specified ranges of the DA curve. While the theoretical discussion of which yield model is physically accurate continues, a model must be chosen for use today. The selection of a 'best' model by statistical methods is done with so called goodness of fit techniques and is discussed in the model section of this paper.