Analysis of Co-Integration of Particular Annual Time-Series Data of Traffic Accidents in the Czech Republic by the Way of Vector Auto-Regression Model (VAR)
Dr. Zdeněk Kovařík, CSc., Department of science and research PA Czech Republic in Prague
pplk. JUDr. Ing. Marek Blažejovský. Ph.D., Faculty PA Security Legislation of the Czech Republic in Prague, Department of Criminal Police

The presented scientific article deals with the testing of co-integration among annual time-series data of traffic accidents, traffic accidents under the influence of alcohol and traffic accidents under the influence of illegal and legal drugs in the Czech Republic. The paper focuses on more advanced analysis of these indicators where the analysis itself is approached in a more complex manner and time relations are analysed interdependently as a one co-integrated system. What is clear from the complex nature of the problem is the fact that the relation among given variables comes into interaction with series of further factors which we are not able fully uncover, let alone integrating them into a reasonable model. Furthermore, the used time series are very short, which constitutes also a certain problem. Nevertheless, the results of the analysis can be considered to be correct and reasonable. For the purposes of prediction the model of vector auto-regression has proven to work effectively. Apart from its heurist aspect the original scientific research paper has a strong didactic dimension. The paper allows the concerned reader to apply analogy that helps to solve similar problems.
Key words: time series, non-stationarity, unit root testing, co-integration, vector auto-regression model (VAR), vector error correction model (VEC), Granger causality, residuals, mean absolute percentage error (MAPE), Thiel’s index.