Vi. Dvorkin, DATA-PROCESSING IN THE INTERLABORATORY TEST BY ANALYSIS OF COVARIANCE, Chemometrics and intelligent laboratory systems, 22(1), 1994, pp. 127-146
The results of an interlaboratory test on the determination of total c
holesterol and triglycerides content in standard reference materials (
frozen human blood sera) are discussed. The test's setup included the
possibility of estimating the concentration-dependent effects bath bet
ween laboratories and within each laboratory (run-to-run effects). The
interlaboratory test data have been calculated on the basis of the mo
del of analysis of covariance providing simultaneous assessment of qua
litative (here, concentration-independent) and quantitative (proportio
nal to concentration) effects. Intra- and interlaboratory effects, bot
h independent on the concentration of the substances measured and thos
e proportional to them, were significantly detected. These effects are
interpreted as constant and proportional constituents of the error at
various levels: between runs of the analysis within each laboratory;
between laboratories; for all laboratories participating in the test.
It was shown that with analysis of covariance it is possible to detect
the errors where they cannot be detected with the conventional method
of the analysis of variance. It is shown that the analysis of covaria
nce is extremely useful to develop a recommendation to reduce intra- a
nd interlaboratory errors of analysis, as it allows the researchers to
assess separately constant and proportional constituents of the error
typically coming from different sources.