Abstract
Graphical models of conditional independences (CIG) are an important instrument of the multivariate statistics. They describe and transparently represent the structure of dependence relationships in a given set of random vectors.
The principal aim of this paper is exploring the possibilities of application data analysis by graphical models to the long period time series of daily aggregated numbers of deaths by individual causes of death and the database of physical parameters of the ionosphere-inner magnetosphere region which can influence human organism. In particular, we focused on the causes of death according to ICD-10 of groups VI.Diseases of the nervous system and IX.Diseases of the circulatory system.
We used time series of daily aggregated numbers of deaths separately for both sexes at the age groups under 39 and over 40+. This method appears to be useful for studying this correlationships and can be applied even in the case when classical parametric methods are not convenient, e.g. for non-continuous time series etc.
We consider the structure of pairwise dependence of its individual components, looking for the maximum likelihood estimate of the variance matrix under conditions given by the graphical model. The CIG method allowed us to implement additional time series variables into model and the data best fit model is computed.
We employ CIG multivariate statistic methods applied also to long period daily observational data for find out intragroup relationships between daily aggregated numbers of deaths by individual causes of death in numerous groups of diagnoses according to ICD-10.