Simplified implementation of the windowing method for systematic and random yield calculation

To separate systematic and random yield loss, common implementations of the windowing method use an unweighted fit of the Poisson model. By comparison to a properly weighted fit of the negative binomial model, we show that the unweighted fit of the Poisson model can give highly inaccurate results, especially in the presence of clustering. The unweighted fit of the Poisson model is shown to improve by reducing the number of points included, with the best results given by the very simplest form, which uses two points only.