Performance of laboratories measuring silica in the Proficiency Analytical Testing program

A statistical study was performed on the results reported by laboratories analyzing silica samples in the first 101 rounds of the Proficiency Analytical Testing (PAT) program. Five laboratories participated in the first round of the PAT program in 1972, and participation grew to 130 laboratories before falling to 105 in Round 101. The laboratories use all three of the major methods of analysis: colorimetry, x-ray diffractometry, and infrared spectroscopy. The objectives of the study were to determine bias between methods, the variability associated with the methods, and any changes in bias or variability caused by a number of factors. The colorimetric method has consistently given the lowest results, particularly at higher loadings. X-ray diffractometry results were biased higher than infrared spectroscopy results during one period but not in the following period. Between the two periods, the procedures and materials used to prepare PAT samples changed in a number of ways, but the switch to quartz dust with a smaller particle size is a likely explanation for the bias difference. Generally, silica analyses have improved in precision over time, and this improvement has taken place for all three of the methods. The colorimetric method has shown the poorest precision of the three methods, but, unlike the differences in bias, the differences in precision have diminished considerably over time. Precision estimates from other studies were compared to those from this study to learn more about sources of variability. The largest source of variability, the differences between laboratories, was large even when laboratories used the same method, as they did in a collaborative study of silica methods.