2018 Volume 22 Issue 2 Pages 76-82
Meat inspection data summarize and integrate the results of meat inspections at slaughterhouses conducted by official veterinarians with local government meat inspection centers. Although meat inspection data has been utilized for surveys and research and fed back to producers to improve hygiene in meat production, few cases have been reported in which anything more than listing or drawing graphs of meat inspection data is performed. However, it is difficult to determine objectively whether livestock breeding conditions tend to become better or worse based on the enumeration of figures or quantity of change in graphs. That is, statistical methods that provide criteria on which the decisions made by producers, veterinarians, and administrators are based are needed. Time series analysis is a suitable method to analyze meat inspection data as it comprises sequential time series data. Since time series analysis involves a variety of methods, to analyze meat inspection data, it is necessary to select a proper method based on the statistical distribution of data. Therefore, the purposes of the present study were to identify the optimal method for using multiple models properly according to the distribution of data and to analyze data with different types of distributions comprehensively.
Investigations were carried out to obtain the favorable conditions for the SARIMA model and two-part model to detect anomalies in the condemnation rate and number. It became clear that the SARIMA model is suitable for analyzing condemnation rate data that do not deviate excessively around 0% or 100%, as the logit-transformed condemnation rate is assumed to be normally distributed. Our results also showed that a two-part model with autoexponential regression can adequately deal with data involving an excessive number of zeros. Therefore, the present study paves the way to enable providing administrative information regarding various diseases based on objective criteria using the SARIMA model for diseases in which the data rarely contain zeros and the two-part model for diseases in which the data contain many zeros.
