Learning Curves and Appropriate Regression Methodologies.

We focus on estimation for two learning curve models of the form Y = aXb and I* = cXd. In the first, or average model, Y is the average cost per unit for the first X units. In the second, or unit model, I is the incremental cost for the Xth unit. The paper addresses the estimation of the parameters a and b, or c and d, when the available data are not suitable for directly fitting the equation Y or I*. It derives the equations of the midpoint methodology of the unit model and proposes a separate midpoint formula for the average model. We conclude with a summary table of data types and corresponding regression equations, and an extensive illustration of the different techniques for pedagogic purposes. These analyses clarity the regression differences between the two learning models, which have often been confused in the literature.