1. Comparing forecast models Aa Aa A statistician for the Boston Police Department developed forecasts of monthly armed robberies in Boston from June 1974 to May 1975 using two forecasting methods, m...
1. Comparing forecast models Aa Aa A statistician for the Boston Police Department developed forecasts of monthly armed robberies in Boston from June 1974 to May 1975 using two forecasting methods, model A and model B. His forecasts, absolute forecast errors, and squared forecast errors for both forecasting techniques, along with the actual time series, are displayed in the following table. The time series plot of the data is shown under the table. (Data source: Time Series Data Library.) Model A Absolute Forecast Error Model B Absolute Forecast Error Month Squared Forecast Forecast Squared Forecast Error Armed Forecast Robberies Error 287 355 460 June 1974 355 364 460 360.60 444.60 367.80 466.20 445.80 397.00 484.20 451.00 390.20 380.20 324.20 31.36 237.16 96.00 96.00 35.00 9216.00 9216.00 1225.00 15.40 September October November 487 452 391 500 451 375 20.80 432.64 38.44 36.00 249.64 452 452 3721.00 January 1975 49.00 2401.00 15.80 451 375 372 355 March April May 15.20 231.04 67.24 492.84 1,830.80 302 53.00 2809.00 22.20 Sum 390.00 28,588.00 119.20 500 400 300 Jun u Aug Sep ct Nov Dec an Feb Ma Apr May 1974 1975 For model A, the mean absolute error (MAE) is For model B, the mean absolute error (MAE) is According to the MAE, provides more accurate forecasts of the number of armed robberies in Boston. For model A, the mean squared error (MSE) is For model B, the mean squared error (MSE) is According to the MSE provides more accurate forecasts of the number of armed robberies in Boston. In a situation where MAE and the MSE provide conflicting guidance on forecast accuracy, the choice of measure is made on the basis of forecast preferences. If avoiding large errors is important, should be used
1. Comparing forecast models Aa Aa A statistician for the Boston Police Department developed forecasts of monthly armed robberies in Boston from June 1974 to May 1975 using two forecasting methods, model A and model B. His forecasts, absolute forecast errors, and squared forecast errors for both forecasting techniques, along with the actual time series, are displayed in the following table. The time series plot of the data is shown under the table. (Data source: Time Series Data Library.) Model A Absolute Forecast Error Model B Absolute Forecast Error Month Squared Forecast Forecast Squared Forecast Error Armed Forecast Robberies Error 287 355 460 June 1974 355 364 460 360.60 444.60 367.80 466.20 445.80 397.00 484.20 451.00 390.20 380.20 324.20 31.36 237.16 96.00 96.00 35.00 9216.00 9216.00 1225.00 15.40 September October November 487 452 391 500 451 375 20.80 432.64 38.44 36.00 249.64 452 452 3721.00 January 1975 49.00 2401.00 15.80 451 375 372 355 March April May 15.20 231.04 67.24 492.84 1,830.80 302 53.00 2809.00 22.20 Sum 390.00 28,588.00 119.20 500 400 300 Jun u Aug Sep ct Nov Dec an Feb Ma Apr May 1974 1975 For model A, the mean absolute error (MAE) is For model B, the mean absolute error (MAE) is According to the MAE, provides more accurate forecasts of the number of armed robberies in Boston. For model A, the mean squared error (MSE) is For model B, the mean squared error (MSE) is According to the MSE provides more accurate forecasts of the number of armed robberies in Boston. In a situation where MAE and the MSE provide conflicting guidance on forecast accuracy, the choice of measure is made on the basis of forecast preferences. If avoiding large errors is important, should be used