Data Science For Professionals: Experimental Designs with Least Resources [Paperback] Prof. N.C. Das Paperback – 12 October 2019

The book is meant for a wide-spectrum of readership-empirical scientists, consultants, technocrats, advertisers, researchers and students learning Data Science for conducting small replicated or non-replicated experiments / demonstrations.

It deals with two-factor three-level designs to be conducted only in six experimental units. For this an exhaustive set of 9c6=9c3 = 84 designs have been generated. Least Square algorithm has been applied for estimating separate effect of each component and their interaction along with fitting of response surface; meeting both the objectives for any factorial experiment. Of them seventy-six could be found as valid or estimable designs and the rest as non-estimable ones. Value of the determinant obtained for Least Square Matrix for each such design has been the indicator of its D-optimality status: the same could also be obtained from geometric structure of the designs. Such exhaustive search ascertains (8+4) D-optimal designs; of them only (1+3) are hitherto known. Also, rotatable feature of any such design does not affect its D-optimality status.

A simple method SAME has been devised to meet the said twin objectives, thus escaping matrix formation and related operations.

Each such design could be combined to form pair of replicates or blocks of an experiment so as to fetch statistical significance test based on ANOVA and fitting the response surface.