Statistical comparison of two laboratory methods, measuring the same objects, is usually performed by regression. However, the results achieved are often incorrect due to application of ordinary least squares linear regression, which should not be used for this purpose. A statistically correct decision whether two laboratory methods provide concordant or discordant results is reliable only when using regression techniques which respect the errors of both compared variables.Another correct possibility is to apply a robust regression where the errors of compared variables are not influencing the calculation of regression parametersand their standard deviations. A survey of appropriate regression techniques and their use for comparison of real-life clinical methods is given in this paper.
eng
dc.format
p. 119-129
cze
dc.language.iso
eng
dc.publisher
Univerzita Pardubice
cze
dc.relation.ispartof
Scientific papers of the University of Pardubice. Series A, Faculty of Chemical technology. 10 (2004)
eng
dc.rights
open access
eng
dc.subject
statistické metody
cze
dc.subject
regrese
cze
dc.subject
srovnávací studie
cze
dc.subject
laboratorní metody
cze
dc.subject
cholesterol
cze
dc.title
Correct ways of using regression for method comparison studies - determination of LDL-cholesterol