In: Statistics and Probability
An intensive care unit manager would like to forecast the number of inpatient admissions for the next year. Inpatient admissions for the unit over the past 10 years are shown in the following table.
Year |
1 |
2 |
3 |
4 |
5 |
6 |
7 |
8 |
9 |
10 |
Admissions |
321 |
345 |
330 |
368 |
405 |
395 |
456 |
469 |
474 |
465 |
Prepare a forecast for admissions for years 4 through 11 by using a three-year moving average method.
We prepare this by first taking total of preceding 3 years of the forecast year(since 3-yer moving average). Then we divide it by 3.
Year | Admissions | 3 moving total | Average |
1 | 321 | ||
2 | 345 | ||
3 | 330 | ||
4 | 368 | 321+345+330 = 996 | 996/3 = 332.000 |
5 | 405 | 1043 | 347.667 |
6 | 395 | 1103 | 367.667 |
7 | 456 | 1168 | 389.333 |
8 | 469 | 1256 | 418.667 |
9 | 474 | 1320 | 440.000 |
10 | 465 | 1399 | 466.333 |
11 | 1408 | 469.333 |
Calculate the mean squared error (MSE) for the three-year moving average method, with error measurement beginning in year 4.
We calculate the error (Actual (A)- forecast (F)) and then square it. We divide this by 7 (Since we we have forecast for 7 years).
Year | Admissions | 3 moving total | Average (F) | Error | Error^2 |
1 | 321 | ||||
2 | 345 | ||||
3 | 330 | ||||
4 | 368 | 996 | 332.000 | 36.000 | 1296.000 |
5 | 405 | 1043 | 347.667 | 57.333 | 3287.111 |
6 | 395 | 1103 | 367.667 | 27.333 | 747.111 |
7 | 456 | 1168 | 389.333 | 66.667 | 4444.444 |
8 | 469 | 1256 | 418.667 | 50.333 | 2533.444 |
9 | 474 | 1320 | 440.000 | 34.000 | 1156.000 |
10 | 465 | 1399 | 466.333 | -1.333 | 1.778 |
Total | 13465.889 |
.Mean squared error = = 13465.889 / 7
Assume that the initial forecast for year 1 is 340 admissions. Prepare a forecast for admissions for years 2 through 11 by using an exponential smoothing method with = 0.30.
We have the following formula for forecasting
Year | Admissions (A) | Forecast (F) |
1 | 321 | 340.000 |
2 | 345 | 334.300 |
3 | 330 | 337.510 |
4 | 368 | 335.257 |
5 | 405 | 345.080 |
6 | 395 | 363.056 |
7 | 456 | 372.639 |
8 | 469 | 397.647 |
9 | 474 | 419.053 |
10 | 465 | 435.537 |
11 | 444.376 |
Calculate the mean squared error (MSE) for the
exponential smoothing method with
a = 0.30, with error measurement beginning in year
4.
It is calculated the same way as moving average.
Year | Admissions (A) | Forecast (F) | Error | Error^2 |
1 | 321 | 340.000 | ||
2 | 345 | 334.300 | ||
3 | 330 | 337.510 | ||
4 | 368 | 335.257 | 32.743 | 1072.104 |
5 | 405 | 345.080 | 59.920 | 3590.418 |
6 | 395 | 363.056 | 31.944 | 1020.424 |
7 | 456 | 372.639 | 83.361 | 6949.031 |
8 | 469 | 397.647 | 71.353 | 5091.193 |
9 | 474 | 419.053 | 54.947 | 3019.153 |
10 | 465 | 435.537 | 29.463 | 868.055 |
Total | 21610.377 |
MSE = 21610.377 / 7
Compare the performance of the two forecasting methods by
using MSE as the performance criterion. Which method would you
recommend based on MSE? Why?
MSE of 3 year moving avg (1923.698) > MSE of exponential smoothing (3087.197)
We can conclude that the 3 -year moving average method is better for forecasting since it produces less error. (The is less variation between the actual and the forecast values)