In routine systems investigating the morbidity according to diagnosis it is very useful to analyse the development in time (for example the development of weekly reports). This paper is concerned with the methodology of such analyses. In practice it appears that the number of cases depends on season. It stands to reason, that it is necessary to consider also long-therm trends. In this paper two different approaches are discussed – the Box-Jenkins analysis, which describes the random error and the Method of Trend Decomposition which spread the number of cases into the systematic component (long term trend and seasonal effect) and random variability. The authors describe the method of smoothing the estimate of the time series by kernel estimate. In both approaches they use weekly reports from the whole Czech Republic of diagnoses viral hepatitis A, rubella and salmonellosis.

        Key words: time series – kernel estimate – analysis of number of cases.