Source (1/20)

Woods, D.C., Lewis, S.M., Dewynne, J.N. "Designing Experiments for Multi-Variable B-Spline Models." Sankhya: The Indian Journal of Statistics 65.3 (2003): 660-677. Print.


For many explanatory variables, different models are derived from the B-spline and monomial basis functions. Efficiency is an important factor when analyzing the various models defined. Woods presents an algorithmic model for designing experiments built from simpler models.First, the article begins by explaining the current modeling techniques, mainly low order polynomials or simple nonlinear functions. Then, using examples, Woods explains why such modeling techniques are not adequate for specific experiments. Drawing on research for different authors, Woods explains how the simple model has been built up into partitioning the polynomial into different splines. However, he shows when applying this idea of polynomial splines, the degrees and the number of knots is unknown, amplifying the complexity of the problem of modeling . After mathematically and statistically proving (complete with examples) how to extend the simple polynomial spline to the more complex class of experiments, Woods applies the mathematical theory to optimal design models. Finally, Woods concludes with two methods to construct part of the algorithm necessary to create the optimal design models.Like the previous paper analyzed, this paper was authored by the professor I will be working with in Southampton. Through reading his published works, I can better understand Dr. Woods’ background and areas of interest. Also, this paper provides me with a better experimental design understanding.