農業経営管理に用いる標準値としての鐘形頻度最頻値の適合性についての検討

平年収量の予測精度からの接近

抄録

Standard values for production indices, such as yield and labor hours, must be set in advance when setting the standard costs for cost control of agricultural products and planning production plans for the PDCA cycle for progress control. Standard production indices are the true values that can be achieved with normal efforts under specific production conditions, excluding random fluctuations. Standard agriculture values are obtained from representative values of past performance data. Production index distribution is not symmetrical and likely to contain outliers. Existing studies have assumed that the mode, which is positioned as typical or normal, is more appropriate than the mean, trimmed mean, and median as the representative value for obtaining standard values. In this study, we adopted the bell-form frequency mode and examined the validity of this assumption based on data. Accordingly, we compared the prediction errors of each representative value for rice and soybean yields for each of the past 30 years with the normal year yield for each year. The predicted values were representative values of the time series data for the past 6, 8, and 10 years, excluding the predicted year. For error measurement, the expected error value （error bias） and the average of the absolute error values (error scatter) were employed. As a result, in the data examined, the error of the mode’s predicted value was often the smallest. Therefore, the mode is a better fit as a standard value than the mean, trimmed mean, and median.