Vi. Dvorkin, CORRECTION TECHNIQUES FOR LONG-TERM STUDIES IN ANALYTICAL-CHEMISTRY, Chemometrics and intelligent laboratory systems, 34(2), 1996, pp. 173-185
There is a new approach to data processing of long-term studies (which
are mostly studies of population samples of specimens) in analytical
chemistry, Specimens of each sample are analyzed in several runs of th
e analysis. In addition to specimens each run analyzed from 2 to 4 ref
erence materials with different concentration of substances to determi
ne and at least two measurements for every material are done. These ma
terials are distributed randomly among specimens under study. The stat
istical models of errors of reference materials study results are used
for the recalibration (corrective action) of population sample study
result, This approach is applied to the determination of lipids (total
cholesterol, triglycerides and cholesterol of high density lipoprotei
ns) in human sera from subjects in the population studies in Tashkent
which made up 12 samples of several hundreds specimens in each. Standa
rd reference materials were the donor's frozen serum certified by refe
rence methods, Data obtained appeared to be well described on the basi
s of linear models of the analysis of covariance and not so well by th
e model of the after-scaling analysis of variance. Correction results
on the basis of both models were compared with the run-wise and sample
-wise result correction, Change in the coefficient of variation of the
sample was considered as the criterion of availability for service an
d effectiveness of the correction. In most cases run-wise calibration
on the basis of either model was the most suitable. The analysis of co
variance model turned out not to be suitable in determining cholestero
l of high density lipoproteins because of unsuccessful study design, O
n the whole, this method proved effective and led to the improvement o
f the population sample studies results. Although corrective action wi
thin after-scaling analysis of variance model is not always as effecti
ve as analysis of covariance model but it is more steady. It can be ex
pected that this approach will prove to be useful in long-term studies
in other areas of analytical chemistry.