Japanese title: Investigate the design caused by DOE custom design platform using JMP Pro 12
University of Central Florida Statistics / Professor Mark Johnson (Mark · Johnson)
School of Science and Technology, Tokyo University of Science Department of Management Engineering Keiichi Yasui
The customized planning platform within DOE of JMP Pro 12 can generate two level executive factor plans of 27 different experiment times 16.In this presentation we will use JMP to investigate the characteristics of various plans and to provide some JSL scripts to facilitate analysis.One of the properties is that in a given plan, a vector consisting of all main effects and two factor interactions (ie, vectors I, A, B, ..., F, AB, AC, ..., EF It is about the space spanned by.Although somewhat surprising, the "Gold Standard" Resolution IV program (generator E = ABC, F = BCD) only spaces 14 dimensions out of the maximum 16 dimensions.This flaw means that the gold standard plan can not model several response vectors (in particular, in this case, they are in the space covered by the vectors ABD and ACD).In contrast, "poor" resolution III plans based on generators E = AB and F = CD can span a 16-dimensional space of possible response vectors, in certain situations, more than the gold standard Are better.Some non-regular plans of 27 possible custom plans also satisfy the full dimension possibilities with properties of other plans (E (s 2) and trace (AA ')).This is as pointed out by Lou Johnson and Anderson Cook (Quality Engineering, 2014).For example, the generators E = ABC and F = 1/2 [CD + ACD + BCD - ABCD] span a 16 - dimensional space.When examining non-regular plans generated from custom plans, it is important to identify all factor placement project plan from the main effect column.JSL scripts are presented to facilitate this effort.Spanning property defects will be shown to occur in the Resolution V plan with 32 experiments.
