Statistics Meeting Book (May 15, 2019)

First page Table of contents Previous page 118 Next page Last page

Conclusions • Analysis of study data using the traditional C metamer method is less desirable unless the analyte is measured in macro concentrations (10% to 100% by composition). Estimates or RSD(r) and RSD(R) are approximations to attempt to approximate the variance sources found more accurately using the pC metamer. • The pC metamer should be first choice for micro or trace analytes. It corrects skewness typically found in the C metamer data. It allows comparison and pooling of variance source estimates (repeatability, laboratory, reproducibility) across concentrations. In the pC metamer, the sizes of these variance sources is the issue, so such indices as HORRAT are not needed. • Outliers should be identified only in the pC metamer, and should be removed only if assignable physical cause is given or the result is so deviant as to be unbelievably homogeneous with the other data. • The normality assumption must always be checked. • If the proper metamer is chosen for a study, RSD(r) and RSD(R) are unnecessary, irrelevant and should not be reported.

Contacts:

• ral@lcfltd.com • qian.graves@fda.hhs.gov

Made with FlippingBook - professional solution for displaying marketing and sales documents online