- 行動計量学 (ISSN:03855481)
- vol.48, no.1, pp.39-48, 2021 (Released:2021-11-10)
In February 2020, Japan began experiencing outbreaks of COVID-19 infection and deaths due to the disease. Open data on the number of positive PCR test results are available from the Ministry of Health, Labour and Welfare (MHLW). We applied the SARIMA models to these time series data and selected an ARIMA(1,1,2) model as the best model among this class of models for the number of infection deaths during the main period, which includes the period covered by the declaration of emergency. We applied bivariate VAR models to the data of the number of PCR-positive persons and the number of infection deaths and selected a VAR(15) model as the best model for the main period. With regards to Granger's causality of the number of PCR-positive persons relative to the number of infected persons and deaths, we obtain clear causality in the data only up until June 17, with causality not so clear after June 18, when MHLW changed the definition of fatalities.