Statistical tools in the clinical laboratory.

Abstract

Method evaluation, control of data and transformation of laboratory results into diagnoses all involve a decision step. A survey of the statistical tools available to organize the information and check the congruity of decision making is provided is focused on: (a) the use of classical statistical tools (including computer based simulation and replication techniques) which enable theoretical distributions to be obtained and their optimal limits to be defined for classification purposes; (b) the analysis of multivariate distributions, which evidences the relationships among the variables involved, whatever they might be: e.g. results obtained on the same specimens with different methods (in test evaluation), different laboratory data related to the same pathophysiological situations (in making diagnoses), etc. As for the latter, the most common techniques of statistical analysis of data (discriminant and cluster analysis, principal components analysis) are also illustrated by general examples.