1st Practical Examination Regression Analysis ( AGST 426)

Time Series Analysis

Name of the Experiment: Freehand method
Fit a trend line to the following data by using the freehand method.
Year: 1991 1992 1993 1994 1995 1996 1997 1998
Sales turnover : (Rs. in lakh): 80 90 92 83 94 99 92 104

Solution:

Sales
120
100
80
60
40
20
0
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

Name of the Experiment: Semi Average Method

Fit a trend line to the following data by the method of semi-average and forecast the sales for
the year 2002.
Year Sales of Firm (thousand units) 1993 1994 1995 1996 1997 1998 1999
Year Sales of Firm (thousand units) 102 108 105 116 114 112 110

Solution: Since number of years are odd in number, therefore divide the data into equal parts
(A and B) of 3 years ignoring the middle year (1996). The average of part A and B is
 Sum of fist 3 years sales 102 + 108 + 105 315
yA = = = = 105
3 3 3
Sum of last 3 years sales 114 + 112 + 110 336
yA = = = = 112
3 3 3

Part A is centred upon 1994 and part B on 1998. Plot points 105 and 112 against their middle
years, 1994 and 1998. By joining these points, we obtain the required trend line as shown Fig.
below. The line can be extended and be used for prediction.
Sales
120

115

110

105

100
1992 1993 1994 1995 1996 1997 1998 1999 2000

Fig. Trend Line by the Method of Semi-Average

To calculate the time-series yˆ = a + bx, we need

y Change in sales 112 − 105 7

Slope b = = = = = 1.75
x Change in years 1998 − 1994 4

Intercept = a = 105 units at 1994

Thus, the trend line is : yˆ = 105 + 1.75x
Since 2002 is 8 year distant from the origin (1994), therefore we have yˆ = 105 + 1.75(8) =
119

Name of the Experiment: Weight Moving Average method

Harpic cleaner sales for 12 months is given below.
Month : 1 2 3 4 5 6 7 8 9 10 11 12
Actual sales (in units) : 10 12 13 16 19 23 26 30 28 18 16 14

Solution:
The owner of the supermarket decides to forecast sales by weighting the past three months as
follows:
3  Last month + 2  Two months ago + 1  Three months ago
Weight Moving Average =
6
th
Weighted Moving Average for 4 Month=
3  Sales in 3rd month + 2  Sales in 2nd month + 1  Sales in 1st month
6
Table 7.3: Weighted Moving Average
Month Actual Sales Three-month Weighted
Moving Average
1 10 -
 2 12 -
3 13 -
4 16 20.17
5 19 23.58
6 23 17.00
7 26 33.50
8 30 39.17
9 28 45.17
10 18 48.17
11 16 38.5
12 14 30.33

3  13 + 2  12 + 1  10 121
Forecast for 3rd Month = = = 20.17
6 6

3  16 + 2  13 + 1 12 141
Forecast for 4th Month = = = 23.58
6 6
Name of the Experiment: Moving Average Method

The data on the rice production during 1980-90 in a certain region are given below :

Year 198 198 198 198 198 198 198 198 198 198 199
0 1 2 3 4 5 6 7 8 9 0
Producti
280 300 325 420 315 360 400 450 350 420 460
on (Ton)

Solution:
Determination of trend by the method of moving average :
It is clear from the data that a 4-year cycle is present. So 4-year moving averages are computed
as shown below :

Year Production (Ton) 4-year 4-year Centralized Moving

Moving Total Moving Average Average
1980 280

1981 300
1325 331.25
1982 325 335.63
1360 340.00
1983 420 347.50
 1420 355.00
1984 315 364.38
1495 373.75
1985 360 377.50
1525 381.25
1986 400 385.63
1560 390.00
1987 450 397.50
1620 405.00
1988 350 412.50
1680 420.00
1989 420
1990 460

The trend line is estimated by plotting the 4-year moving averages along the y-axis against the
corresponding year (plotted along the x-axis) as shown below :

Name of the Experiment: Least Square Method

Estimation of trend of 7 years rice production in a certain district by the method of least
squares-

Year Production (y) Trend value

t t2 yt ŷ = 99 + 2t
(t) in ton
1980 87 -3 9 -261 93
1981 98 -2 4 -196 95
1982 104 -1 1 -104 97
1983 95 0 0 0 99
1984 101 1 1 101 101
1985 108 2 4 216 103
1986 100 3 9 300 105
Total 693 Σt = 0 Σt = 28
2
Σyt = 56

y = a + bt

By OLS method
yt
yt −
b̂ = 7 = 56 − 0 = 2
(t ) 2 28 − 0
t 2 −
7
693
â = y − bt = − 0 = 99
7
  Trend line is ŷ = 99 + 2t.

120

110
Trend value

100

90

80
1980 1981 1982 1983 1984 1985 1986

Year
Fig. 16.5: Least Squares Trend Line.

Name of the Experiment: Method of Simple Average

Computation of Seasonal Variation And Seasonal Index on Quarterly Basis (Price data for four
years of a certain commodity) :
Year 1st quarter 2nd quarter 3rd quarter 4th quarter
1991 50.3 54.8 56.0 58.0
1992 50.1 55.1 55.4 59.5
1993 65.2 69.1 70.0 70.5
1994 67.4 70.0 70.2 72.0
Total 233.0 249.0 251.6 260.0
Average 58.25 62.25 62.90 65.0
Seasonal
93.80 100.24 101.29 104.67
Index

General mean = ¼(58.25 + 62.25 + 62.90 + 65.0) = 62.1

58 .25
Seasonal Index for the 1st quarter = x 100 = 93.80
62 .1
62 .25
Seasonal Index for the 2nd quarter = x 100 = 100.24
62 .1
62 .90
Seasonal Index for the 3rd quarter = x 100 = 101.29
62 .1
65 .0
Seasonal Index for the 4th quarter = x 100 = 104.67
62 .1
Name of the Experiment: Ratio to Trend Method
Quarterly production of a certain product (ton) in a factory during 1984-1987 are given below
:

Year 1st quarter 2nd quarter 3rd quarter 4th quarter

Q1 Q2 Q3 Q4
1984 62 58 70 60
1985 66 60 72 64
1986 64 56 74 70
1987 66 64 74 68

Computation of seasonal variation and seasonal index are described below :

Step-1 : Estimation of the trend line and rate of change

Total production Quarterly average Time Trend value
Year t2 ty
(ton) y t ŷ = â + b̂t
1984 250 62.5 1 1 62.5 62.95
1985 262 65.5 2 4 131.0 64.65
1986 264 66.0 3 9 198.0 66.35
1987 272 68.0 4 16 272.0 68.05

Σy = 262 Σt = 10 Σt2 = 30 Σty = 663.5

t y 10  262
ty − 663 .5 −
4 = 4 8.5
b̂ = = = 1.70
2 (t ) (10 )
2 2 5
t − 30 −
4 4
262 10
â = y − bt = − 1.70  = 65 .5 − 4.25 = 61 .25
4 4
Trend line, ŷ = 61.25 + 1.70 t
1.70
b̂ = 1.72  Quarterly increase = = 0.43
4

Step-2 :
Computation of Quarterly Trend Values for Different Years :
 Quarterly rate of increase is 0.43; 1984 average is 62.5
0.43
 The 3rd quarter trend for 1984 is 62.5 + = 62.715
2
0.43
The 4th quarter trend for 1984 is 62.715 + = 63.145
2
0.43
The 2nd quarter trend for 1984 is 62.5 - = 62.285
2
0.43
The 1st quarter trend for 1984 is 62.285 - = 62.07
2
In the same way for the year 1985, the 1st, 2nd, 3rd and 4th quarter trends are 64.855,
65.285, 66.175 and 66.144 respectively;
Those for the year 1986 are 65.355, 65.785, 66.125 and 66.645 respectively, and
Those for 1987 are respectively 67.355, 67.783, 68.215 and 68.645.
Seasonal variations will be obtained by subtracting the general mean of the quarterly
averages from the corresponding trend values. The quarterly averages are 64.855, 65.285,
66.715 and 66.145 and the general mean is 65.5

Seasonal variations are -0.645, -0.215, 0.645, 0.645

Step-3 :
Computation of quarterly trend-free seasonal indices by expressing the corresponding
value as percentage of the trend value :
For example,
62
Trend free index of the 1st quarter of 1984 is x 100 = 101.52
61 .07
56
Trend free index of the 2nd quarter of 1984 is x 100 = 93.12
62 .285
and so on.
Trend values and trend free seasonal indices (%) for different quarters during the reference
period are presented below :
Trend value Trend free seasonal index (%)
Year
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
1984 62.07 62.285 62.715 63.145 101.52 93.12 111.62 95.02

1985 64.855 65.285 65.715 66.145 101.77 91.90 109.65 96.77

1986 65.355 65.785 66.215 66.645 97.93 85.13 11.76 105.03

1987 67.355 64.785 68.215 68.645 97.99 94.42 108.81 99.00

Average 64.855 65.285 65.715 66.145 99.48 91.14 110.44 98.96

 Since there is no extreme value in the trend free values, seasonal indices for the period 1984-
1987 are computed using the arithmetic mean. As the indices of the four quarters add to 400.02
 400, use of correction term is not required.

Name of the Experiment: Ratio to Moving Average

In a small dairy farm, the quarterly milk production (1000 litres) during the years from 1991 to
1994 are given below :
Year 1st quarter 2nd quarter 3rd quarter 4th quarter
Q1 Q2 Q3 Q4

1991 17 18 30 21

1992 28 22 39 28

1993 31 33 45 47

1994 44 39 47 49

Computation of Seasonal Variation and Seasonal Index by the Method of Ratio to Moving
Average :
First, four quarter moving averages are computed and centralized differences of the
corresponding given values from these centralized, moving averages are the short term
variations. If these centralized moving averages are expressed as percentages of the
corresponding given values, we get the corresponding ratio to moving average.

Computation of Quarterly Moving Average, Short Term Variation and Ratio to Moving
Average :
Ratio to Moving
Value 4 Quarter Moving Centralized Quarterly Short-term variation x
Year Quarter Moving Average ( x i ) ( xi − x ) Average  100
(x) Average x i

Q1 17
Q2 18
21.50
1991 Q3 30 22.875 2.875 131.148
24.25
Q4 21 24.75 3.75 84.848
25.25
Q1 28 26.375 -1.625 104.673
27.50
1992
Q2 22 28.375 6.375 77.533
29.25
 Q3 39 29.625 -9.375 131.646
30.00
Q4 28 31.375 3.375 88.189
32.75
Q1 31 33.500 2.50 92.537
34.25
Q2 33 35.375 2.375 93.286
36.50
1993
Q3 45 38.125 -6.875 118.033
39.75
Q4 37 35.375 -1.625 104.594
31.00
Q1 44 36.375 -7.625 120.962
41.75
Q2 39 43.25 4.25 90.173
1994
44.75
Q3 47
Q4 49
Difference between the general mean of quarterly averages and quarterly averages give the
seasonal variation.

Computation of Seasonal Variation :

Short term variation (Quarterly)
Year
Q1 Q2 Q3 Q4
1991 - - 2.875 3.75
1992 -1.625 6.375 -9.375 3.375
1993 2.50 2.375 -6.875 -1.625
1994 -7.625 4.25 - -
Quarterly
Average ( x i ) -2.25 4.3333 -4.458 1.833

Seasonal variation
(x i − x) -2.1165 4.4668 -4.3245 1.9665

x
=0.1335

Quarterly averages of the ratio to moving averages and the general mean of the quarterly
averages are computed. Seasonal indices are the ratio of these quarterly averages to their
general mean expressed in percentage :

Computation of Seasonal Index :

Ratio to Moving Average (Quarterly)
Year
Q1 Q2 Q3 Q4
1991 - - 131.148 84.848
 1992 104.673 77.533 131.646 88.189
1993 92.537 93.286 118.033 104.594
1994 120.962 90.173 - -
Quarterly
Average ( x i )' 106.057 86.997 126.942 92.5444
Seasonal index
x i 102.83 84.35 123.08 89.73
 100
x
General Mean
x  = 103 .135
As the seasonal indices sum to 399.99 ≈ 400, use of correction factor is not necessary.

Name of the Experiment: Link Relative Method

Quarterly prices of a commodity for five years (1987 to 1991) are given below :
Quarterly Average Price
Year
Q1 Q2 Q3 Q4
1987 25 21 17 26
1988 30 23 17 31
1989 26 24 23 27
1990 26 26 20 30
1991 29 31 21 28

Seasonal variations and seasonal indices are to be computed using the Link relative method.
Solution :
First the link relative (LR) of each quarter as percentage of the preceding quarter is computed
and their quarterly average (ALR) are obtained :
21
LR for Q2 for 1987 =  100 = 84.00
25
17
LR for Q3 for 1987 = 100 = 80.95
21
26
LR for Q4 for 1987 =  100 = 152.95
17
30
LR for Q1 for 1988 =  100 = 115.38 and so on.
26
Calculated link relative (LR) and average link relative (ALR)
Quarterly LR
Year
Q1 Q2 Q3 Q4
1987 - 84.00 80.95 152.95
1988 115.38 76.67 73.91 182.35
1989 83.87 92.31 95.83 117.39
1990 96.30 100.00 67.79 150.00
1991 96.67 106.90 67.74 133.33
Total link relative
Average link 98.06 91.98 79.07 147.20
relative (ALR)
Chain 100 91.98 72.73 107.6
Relative(CR)
Corrected CR 100.00 90.735 70.24 103.325
Seasonal variation 0.089 -0.003 -0.208 0.122
Seasonal Index 109.80 99.63 77.12 118.14

Now assuming CR for Q1 = 100, we compute the CR for Q2, Q3 and Q4 :

ALR of Q 2  CR for Q1 91 .98  100
CR for Q2 = = = 91.98
100 100
ALR of Q 3  CR for Q 2 79 .07  91 .88
CR for Q3 = = = 72.73
100 100
ALR of Q 4  CR for Q 3 147 .20  72 .3
CR for Q4 = = = 107.06
100 100

In order to compute the correction factor, we compute

ALR of Q1  CR for Q 4 98 .06  107 .06

New CR for Q1= = =104.98
100 100
104 .98 − 100
 Correction factor dq = = 1.245
4
 Corrected CR for Q2 = 91.98 - 1.1245 = 90.735
Corrected CR for Q3 = 72.73 - 2 x 1.245 = 70.24
Corrected CR for Q4 = 107.06 - 3 x 1.245 = 103.325
 ACR = ¼(100 + 90.735 + 70.24 + 103.325) = 91.075

Seasonal variation and seasonal indices are obtained as

CR − ACR
Seasonal variation =
100
CR
Seasonal index =  100
ACR