Wednesday, April 13, 2016: 9:06 AM
335A (Hilton Americas - Houston)
In Dow it is common to manufacture the same product at multiple production sites across the globe. Measurement systems of laboratories at different sites are expected to obtain similar results. One way to assess laboratory performance is to conduct an Inter-laboratory Uniformity (ILU) study. Typical ILU studies have only one factor of interest: laboratory performance. Laboratories in different locations are evaluated to see how they differ in their measurement of a single sample. In this presentation, we will share a form of ILU study where we are interested in evaluating the measurement systems used to count defective particles in a powder at three different laboratories. The number of observations in the ILU study exceeded 400; traditional ILU studies in Dow usually have fewer than 100 observations. The ILU study had 3 factors that were investigated (site, analyst, and operating condition) and many challenges that had to be overcome. The distance between laboratories prevented the use of a fully randomized design. It is infeasible to create homogeneous aliquots of the same material when the variable of interest is a count of low-level defects, yet sample handling had to be minimized to prevent adulteration of samples. In this talk, we will share how we used a split-plot design and non-traditional design elements to overcome these challenges and investigate hypotheses.