Joint Research Conference
June 24-26, 2014
Thermal insulation Standard Reference Materials (SRMs), in this case expanded polystyrene board, are typically used to check guarded-
hot-plate apparatus, calibrate heat-fow-meter apparatus and for checking or calibrating hot-box apparatus, which are all means to measure the thermal conductivity of a material. The thermal conductivity measurements that serve as the basis for the certi ed values for this material were made at the National Institute of Standards and Technology using the guarded-hot-plate method. The thermal conductivity of thermal insulation is typically characterized as a function of bulk density and temperature. So the certi cate for this material contains a certi ed equation, as opposed to a certi ed value. The equation is a polynomial with an intercept and linear terms for both temperature and bulk density. The three coe cients were estimated from data. However, approaches to estimation such as ordinary least squares are inappropriate because the temperature and bulk density for a given measurement of thermal conductivity are themselves measurements and so not known without error. The equation parameters are thus estimated using an errors in variables model under the Bayesian inference paradigm by Markov Chain Monte Carlo simulation. The expanded uncertainty in predictions from this equation is stated as a single relative uncertainty (percentage). It is the maximum relative half width (relative to the prediction) of a symmetric approximately 95% posterior predictive interval, where the maximum is taken over a regular grid of points covering the portion of the covariate space (temperature bulk density) where observations were made.