Algorithm and software for yield analysis

Yield analysis is a method designed to estimate the inference coefficients on each unit of the population, as opposed to regression and other techniques that estimate the inference for the population as a whole. Yield analysis has a great analytical potential, which has been proven in several marketing case studies for large companies. It allows for creation of good models even in situations when traditional analysis fails (e,g. when all correlations are very weak or there is a high multicollinearity). Technically, it could be rooted in random regression coefficients models by Hildreth and Houck (1964), but goes much further. The heuristic algorithm of individual “yields” estimation is presented, which is based on Monte Carlo simulation and bears some convenient analytical features. It is fully controllable by the user. Some details of SAS® code are given; the possibility of using this software for practical application is discussed.