An alternate analysis of a 2x2 DOE using a contour plot of Weibull joint parameter confidence regions

Abstract

Before beginning an accelerated test program for implantable cardiac defibrillator/pacemaker leads, an initial phase of testing was performed to determine if test stations and operators could be used interchangeably. DOE and Monte Carlo simulations were used to set up the 2/spl times/2 test matrix and determine if the test would produce sufficient data. The Weibull analysis methods, including the contour plot, proved to be useful for analyzing the fatigue failure data as well as providing some insight to the analysis methods. The test results demonstrated no difference between the four runs at the 85% confidence level; however, one run was subtly different from the others. A subsequent investigation into the combined data indicated that the three parameter Weibull distribution is more suitable than the two parameter Weibull distribution. The two parameter Weibull must be used for the individual runs, due to the smaller samples sizes, even though the three parameter distribution may represent the best fit to the total population. The observed run-to-run variation may partially be due to the choice of the simpler two parameter Weibull distribution. It is concluded that the effects of test station and operator on Weibull parameters /spl beta/ and /spl eta/, are not significant at the 85% confidence level (this conclusion also holds for higher confidence levels because the intersecting area increases as the confidence level increases). It is also concluded that the three parameter Weibull model yields a better fit than the two parameter Weibull model, for this particular lead specimen, this test method, and when the sample size is sufficient.