1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
|
# PS10.5b, for the AR(24) estimation in Application Section 10.6
open data10-2
genr load1 = load(-1)
# suppress the first observation
smpl 1.02 31.24
# Estimate model with AR(24) using Generalized CORC
ar 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 ; \
load 0 load1 temp
# eliminate all insignificant AR terms with p-value > 0.5
ar 1 2 5 6 7 8 9 12 13 14 15 17 18 19 21 22 23 24 ; \
load 0 load1 temp
# eliminate insignificant AR terms one at a time
ar 1 2 5 6 7 9 12 13 14 15 17 18 19 21 22 23 24 ; \
load 0 load1 temp
ar 1 2 5 7 9 12 13 14 15 17 18 19 21 22 23 24 ; \
load 0 load1 temp
ar 1 2 5 7 9 12 13 14 15 17 19 21 22 23 24 ; \
load 0 load1 temp
# generate forecasts after incorporating the error structure
fcast 2.02 31.24 yf
# Compute forecast error statistics
genr abserror=abs($uhat)
genr abspcter=100*abserror/load
# Flag observations with percent error > 5
genr five =abspcter>5
genr sumfive = sum(five)
# Flag observations with percent error > 10
genr ten = abspcter>10
genr sumten = sum(ten)
# Compute mean absolute percent error
genr mape = mean(abspcter)
print sumfive sumten mape
|
