Statistical Engineering Division Seminar

Joanne Wendelberger
Statistical Sciences Group
Los Alamos National Laboratory
Statistical Engineering Division Seminar
Monday March 13, 2006, 10:30-11:30 AM
NIST North Room 618

Abstract

In the calibration of a mass spectrometer, m specimens of a pure gas are first measured. For each specimen, values are obtained for pressure (P), temperature (T), and an integral of current over time (A). This produces m data vectors from which device sensitivities can be calculated. Inspection of the formula for device sensitivities leads to a fundamental ratio that simplifies further calculations. Techniques for quantification of the precision of estimated device sensitivities and estimation of variance components are described. Propagation of error and Bayesian analysis methods are then proposed to assess precision in the estimation of a mole fraction from a mixture of gases. An approach for ongoing process monitoring is discussed.