Three-decade-old research suggests that although regression coefficients obtained with ordinary least squares (OLS) are optimal for fitting a model to a sample, unless the N over which the model was estimated is large, they are generally not very much superior and frequently inferior to equal...