Methods for evaluating of discrepancy between regularities systems in different groups.
Abstract: A new method of data analysis is discussed. Goal of represented techniques is complete and statistically valid comparing of regularities existing in two different groups of objects. It is supposed that regularities that tie levels of forecasted and explanatory variables are searched with the help of optimal partitioning technique. The developed technique was applied for analysis of genetic factors impact on severity of discirculatory encephalopathy (DEP). At the first stage computer method for evaluating of DEP severity was developed with the help of pattern recognition techniques. It was revealed that computer estimates of severity rather strongly correlate with DD variant of gene coding angiotensin-converting enzyme (ACE). Systems of regularities that ties various clinical or biochemical factors with computer estimates of severity were found with the help of optimal valid partitioning ( OVP) method in groups of patients with three different variants of gene coding ACE. Statistically significant discrepancies were found between revealed regularities systems with the help of developed methods of comparing. Keywords: Optimal partitioning, statistical validity, permutation test, regularities, explanatory variables effect,pattern recognition, discirculatory encephalopathy, genetic factors ACM Classification Keywords: H.2.8 Database Applications - Data mining, G.3 Probability and Statistics -Nonparametric statistics, Probabilistic algorithms