Scribd
Upload
162 views

Time Series hw5

1) The document provides observations from a time series model fitted by zt - 1.1at-1 + 0.28at-2 from periods z91 to z100. 2) It asks to generate 12-period ahead forecasts from period z100 and calculate 80% probability limits. The forecasts are all equal to 168.83 and probability limits are provided from periods z101 to z112. 3) It also asks to generate quarterly sales forecasts for the next 4 quarters using the same model and data, and calculate 80% probability limits for these forecasts. 4) Given a new observation z101=174, it asks to update the forecasts for the next 11 periods using the updating formula

Uploaded by

euler96
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
162 views

Time Series hw5

1) The document provides observations from a time series model fitted by zt - 1.1at-1 + 0.28at-2 from periods z91 to z100. 2) It asks to generate 12-period ahead forecasts from period z100 and calculate 80% probability limits. The forecasts are all equal to 168.83 and probability limits are provided from periods z101 to z112. 3) It also asks to generate quarterly sales forecasts for the next 4 quarters using the same model and data, and calculate 80% probability limits for these forecasts. 4) Given a new observation z101=174, it asks to update the forecasts for the next 11 periods using the updating formula

Uploaded by

euler96
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

1

時間序列作業 5 08/04/14

5.2 The following observations represent the values z91 , z92 ,..., z100 form a series
fitted by the model zt  at  1.1at 1  0.28at 2
166,172,172,169,164,168,171,167,168,172.
(a) Generate the forecasts zˆ100 ( l ) for l  1, 2,...,12 and draw a graph of the series
values and forecasts (assume a90  0, a91  0).
(b) With ˆ a2  1.103, calculate the estimated standard deviations ˆ ( l ) of the
forecast errors and use them to calculate 80% probability limits for the forecasts.
Insert these probsbility limits on the graph, on either side of the forecasts.

a)
 pd q

zˆt (l )  E    j zˆt (l  j )   j at l  j 
 j 1 j 1 
 model : zt  at  1.1at 1  0.28at 2 (ie: ARIMA(0,1,2))
zˆt (l )  E 1 zˆt (l  1)  1at l 1   2at l 2  .
 the forecasts are following:
zˆ100 (1)  z100  1.1a100  0.28a99  167.79
zˆ100 (2)  zˆ100 (1)  0.28a100  168.83
zˆ100 (2)  zˆ100 (3)  ...  zˆ100 (12)  168.83.

b)
note 1: 1   probaility limits zt l ( ) and zt 1 (  ) given by
1/ 2
 l 1

zt l (  )  zˆt (l )  u / 2  1   2j  sa .
 j 1 
note 2: Suppose U ~ N (0,1).
note 3:  j  1 j 1   2 j 2  ...   p  d j  p d   j for j  0
with  0  1,  j  0 for j  0 and  j  0 for j  q.
zt  (1  B  B 2  ...)(at  1.1at 1  0.28at 2 )

 at  0.1at 1   0.18at  j
j 2

80% probability limits for forecasts


j zj(±) j zj(±)
101 (166.446, 169.134) 107 (167.379, 170.290)
102 (167.484, 170.186) 108 (167.360, 170.310)
103 (167.462, 170.207) 109 (167.340, 170.330)
104 (167.441, 170.228) 110 (167.320, 170.349)
105 (167.420, 170.249) 111 (167.301, 170.369)
106 (167.400, 170.270) 112 (167.282, 170.387)
2
時間序列作業 5 08/04/14

5.3 Suppose that the data of Exercise 5.2 represent monthly sales.
(a) Calculate the minimum mean square error forecasts for quarterly sales for
1, 2, 3, 4 quarters ahead, using the data up to t  100.
(b) Calculate 80% probability limits for these forecasts.
a)
forecasts Data
z97, z98, z99, z100 z90, z91, z92, z93, z94, z95, z96

zˆ96 (1)  z96  1.1a96  0.28a95  168.124 Residual of z_97 ~ z_100

zˆ96 (2)  zˆ96 (1)  0.28a96  167.683


4

zˆ96 (2)  zˆ96 (3)  zˆ96 (4).


2

e96 (1)  z97  zˆ96 (1)  2.876


e96 (2)  z98  zˆ96 (2)  0.683
e_96

e96 (3)  z99  zˆ96 (3)  0.318


-2

e96 (4)  z100  zˆ96 (4)  4.318


M
1
e
-4

 MSE  2
l  6.869
M l 1 167.80 167.90 168.00 168.10

b) x_96

z97 (  )  zˆ96 (1)  1.28 Var[e96 (1)]  (1.876, 3.876)


z98 (  )  zˆ96 (2)  1.28 Var[e96 (2)]  (-1.817, 0.203)
z99 (  )  zˆ96 (3)  1.28 Var[e96 (3)]  (-0.849, 1.236)
z100 (  )  zˆ96 (4)  1.28 Var[e96 (4)]  (3.119, 5.268)
3
時間序列作業 5 08/04/14
5.4 Using the data of Exercise 5.2, and given the further observation z101  174
(a) Calculate the forecasts zˆ101 ( l ) for l  1, 2,...,11 using the updating formula
zˆt 1 (l )  zˆt (l  1)   l at 1
(b) Verify these forecasts using the difference equation directly.

a)
zt  at  1.1at 1  0.28at 2
a101  z101  z100  1.1a100  0.28a99  6.21
zˆt 1 (l )  zˆt (l  1)   l at 1
zˆ101 (1)  zˆ100 (2)   1a101  168.83  0.1  6.21  168.21
zˆ101 (2)  zˆ100 (3)   2a101  168.83  0.18  6.21  169.95
zˆ101 (2)  zˆ101 (3)  ...  zˆ101 (11)  169.95.

b) By defnition
zt 1  zt  at 1  1.1at  0.28at 1
pd q
zˆt 1 (l )    j zˆt 1 (l  j )   j at 1l  j
j 1 j 1

 zˆ101 (1)  z101  1.1a101  0.28a100  174  1.1  6.21  0.28  3.73  168.21
zˆ101 (2)  zˆ101 (1)  0.28a101  169.95
zˆ101 (l )  zˆ101 (l  1) for l  3

5.5 For the model zt  at  1.1at 1  0.28at -2 of Exercise 5.2


(a) Write down expressions for the forecast errors et (1),..., et (6), form the same origin t.
(b) Calculate and plot the autocorrelations of the series of forecast errors et (3).
(c) Calculate and plot the correlations between et (2) and et ( j ) for j  1...6.
a)

zt  at  0.1at 1   0.18at  j (ie:  0  1,  1  0.1,  j  0.18 for j  2)
j 2

et (l )  at l   1at l 1   2at l 2  ...   l 1at 1


 at l  0.1at l 1   j 2 0.18at l  j
l 1

b)

zt  at  0.1at 1   0.18at  j (ie:  0  1,  1  0.1,  j  0.18 for j  2)
j 2

  l 1 i i  j
 i  jl 1 0 jl
 [et (l ), et  j (l )]     i2
i 0

 0 jl
4
時間序列作業 5 08/04/14

j ρ[et(3), et-j(3)]
0 1.000
1 -0.113
2 0.173
o.w. 0.000

c)


l 1
 i j i
 [et (l ), et (l  j )]  i 0

 
12
 h0 h2  g 0  g2
l 1 l  j 1

j ρ[et(2), et(j)]
1 -0.100
2 1.000
3 -0.115
4 0.155
5 0.153
6 0.151

You might also like