TIME SERIES ANALYSIS

Lecture # 13

Dr. Imran Khalil

[email protected]
 Contents
• Introduction
• Components of a Time Series
• Secular Trend
• Seasonal Variations
• Cyclical Variations
• Irregular or Random Variations

2
 Time Series Analysis
• A time series consists of numerical data collected, observed or
recorded at more or less regular intervals of time each hour, day,
month, quarter, or year.
• The analysis of a time series is a process by which a set of
observations in a time series is analyzed.
• Examples:
• Weather data
• Stock market prices
• Sales data
• Website traffic
• Energy consumption
• Population growth, etc. 3
 Components of Time Series
1. Secular Trend (T)
2. Seasonal Variation (S)
3. Cyclical Variation (C)
4. Irregular or random variations (I)

4
 Components of Time Series
1. Secular Trend (T)
• A secular trend represents the long-term movement or pattern in
a time series data that shows consistent growth, decline, or
stability over an extended period, typically ignoring short-term
fluctuations.
• Examples:
• Population Growth
• Economic Growth
• Technological Advancement
• Education level
• Climate change, etc. 5
Components of Time Series

6
 Components of Time Series
2. Seasonal Variations (S)
• Seasonal variations are short-term, regular, and repetitive
patterns in time series data that occur at specific intervals due to
seasonal factors like weather, holidays, or cultural events. These
patterns repeat consistently over time, such as daily, weekly,
monthly, or annually.
• Examples:
• Retail sales
• Tourism
• Electricity usage
• Agricultural production
• Hospital visits, etc. 7
Components of Time Series

8
 Components of Time Series
3. Cyclical Fluctuations (C)
• Cyclical fluctuations are long-term, non-regular variations in time
series data that occur due to changes in the economic, social, or
business environment. These fluctuations typically follow cycles
of expansion and contraction but do not have a fixed or
predictable duration, unlike seasonal variations.
• Examples:
• Economic recessions and booms
• Real estate market
• Stock market cycles
• Unemployment rates
• Business Investments, etc. 9
Components of Time Series

10
 Components of Time Series
4. Irregular or Random Variations (I)
• Irregular or Random Variations in time series refer to unpredictable
fluctuations caused by unforeseen or extraordinary events that cannot
be explained by trends, seasonal patterns, or cycles. These variations
are typically short-term and have no identifiable repetitive pattern.
They are often considered "noise" in data analysis.
• Examples:
• Natural disasters
• Economic stocks
• Pandemics
• Technological breakthroughs
• Political events, etc. 11
Components of Time Series

12
 Analyzing Secular Trend
1. The method of freehand curve
2. The method of semi average
3. The method of moving average
4. The method of least square

13
 Secular Trend - Freehand Curve
• Plot the given data on a graph and all points are joined together
with a straight line.
• Observe the up and down movements on the graph and draw a
smooth curve or a straight line freehand passing through the
plotted points in a way such that the general direction of change in
values is indicated.
• This method is simple and quick.
• Trends depends on individual judgement and experience.

14
 Secular Trend - Semi Average
• In this method, we classify the time series data into two equal
parts and then calculate averages of each half.
• If data are even, it is easily divided into two parts.
• If data are odd, then the year at the middle of the time series is left
and then two halves are constituted with the period on each side
of the mid year.

15
 Secular Trend - Semi Average - Example
Find the semi average trend of the following series:

Year Profit Total Average

1973 85
1974 97
1975 100
1976 90
1977 83
1978 105
1979 112
1980 120
16
 Secular Trend - Semi Average - Example
Find the semi average trend of the following series:

Year Profit Total Average

1973 85
1974 97
372 93
1975 100
1976 90
1977 83
1978 105
420 105
1979 112
1980 120
17
Secular Trend - Semi Average - Example

18
 Secular Trend - Semi Average - Example
Find the semi average trend of the following series:
Year Patients Total Average
1967 276
1968 270
1969 260
1970 286
1971 302
1972 321
1973 351
1974 348
1975 346 19
 Secular Trend - Semi Average - Example
Find the semi average trend of the following series:
Year Patients Total Average
1967 276
1968 270
1092 273
1969 260
1970 286
1971 302
1972 321
1973 351
1366 341.5
1974 348
1975 346 20
Secular Trend - Semi Average - Example

21
 Secular Trend - Moving Average
• The 𝑘 period moving averages are defined as the averages
calculated by using the 𝑘 consecutive values of the observed
series, then repeating the operation by dropping one value at the
beginning and including the first value after the preceding total,
and so on, 𝑚𝑜𝑣𝑖𝑛𝑔 𝑜𝑛 one value to calculate each successive
average. This process is continued till the last 𝑘 consecutive values
have been averaged.
𝑘 𝑘+1 𝑘+2
1 1 1
𝑎1 = ෍ 𝑌𝑡 , 𝑎2 = ෍ 𝑌𝑡 , 𝑎3 = ෍ 𝑌𝑡 , …
𝑘 𝑘 𝑘
𝑡=1 𝑡=2 𝑡=3
22
Compute 3-years, 5-years, and 7-years moving averages:
Year Values 3 Year Moving 5 Year Moving 7 Year Moving
Total Average Total Average Total Average
1918 18.0 … … … … … …
1919 20.5 58.1 19.4 … … … …
1920 19.6 64.3 21.4 110.1 22.0 … …
1921 24.2 71.6 23.9 117.2 23.4 161.1 23.0
1922 27.8 77.1 25.7 122.6 24.5 173.3 24.8
1923 25.1 78.8 26.3 133.2 26.6 186.8 26.7
1924 25.9 81.2 27.1 143.0 28.6 203.2 29.0
1925 30.2 90.1 30.0 151.2 30.2 214.0 30.6
1926 34.0 100.2 33.4 161.1 32.2 222.2 31.7
1927 36.0 105.0 35.0 171.0 34.2 237.8 34.0
1928 35.0 106.8 35.6 181.7 36.3 260.3 37.2
1929 35.8 111.7 37.2 196.1 39.2 … …
1930 40.9 125.1 41.7 … … … …
1931 48.4 … … … … … 23 …
Secular Trend - Moving Average - Example

24
Secular Trend - Moving Average - Example

25
Secular Trend - Moving Average - Example

26
 Year Values 4-quarter moving 4-quarter centered 4-quarter centered
(1) (2) total (3) moving totals (4) moving averages (4)/8
1979-I 72
4-Quarter centered Moving Average

II 98
III 79
IV 106
1980-I 79
II 122
III 101
IV 143
1981-I 94
II 141
III 128
IV 160
1982-I 125
II 143
III 135
27
IV 187
Secular Trend - Moving Average - Example

28
 Secular Trend - Method of Least Square
• The Least Squares Method in time series analysis is a
mathematical approach used to fit a trend line to observed data
points. The primary goal is to minimize the sum of the squared
differences (errors) between the observed values and the values
predicted by the trend line. This trend line can then be used for
forecasting or understanding the overall direction of the data.
𝑌෡𝑡 = 𝑎 + 𝑏𝑥
σ𝑌
𝑎=
𝑛
σ 𝑋𝑌
𝑏=
σ 𝑋2 29
 Secular Trend - Least Square - Example
Determine the trend line by the least-squares method from the
following data. Plot the actual values and the linear trend on the
same graph. 𝒀𝒆𝒂𝒓 (𝒕) 𝒀
1945 3
1946 6
1947 2
1948 10
1949 7
1950 9
1951 14
1952 12
1953 18
30
 Secular Trend - Least Square - Example
𝒀𝒆𝒂𝒓 (𝒕) 𝑪𝒐𝒅𝒆𝒅 𝒀𝒆𝒂𝒓 𝒀 𝑿𝒀 𝑿𝟐 𝑻𝒓𝒆𝒏𝒅
(𝑿) 𝒀𝒕 = 𝟗 + 𝟏. 𝟕𝑿
1945 -4 3 -12 16 2.2 σ 𝑌 81
𝑎= = =9
1946 -3 6 -18 9 3.9 𝑛 9
1947 -2 2 -4 4 5.6 σ 𝑋𝑌 101
𝑏= 2
= = 1.7
1948 -1 10 -10 1 7.3 σ𝑋 60
1949 0 7 0 0 9
1950 1 9 9 1 10.7 𝑌෡𝑡 = 𝑎 + 𝑏𝑥
1951 2 14 28 4 12.4 ෢𝒕 = 𝟗 + 𝟏. 𝟕𝑿
𝒀
1952 3 12 36 9 14.1
1953 4 18 72 16 15.8

෍ 0 81 101 60 81

31
Secular Trend - Least Square - Example

32
 Year Coded Year (X) Y XY 𝑋2 Trend
𝑌෡𝑡 = 119.56 + 2.76𝑋
1979-I -15 72 -1080 225 78.2
II -13 98 -1274 169 83.7
III -11 79 -869 121 89.2
Method of Least Square

IV -9 106 -954 81 94.7

1980-I -7 79 -553 49 100.2
II -5 122 -610 25 105.8
III -3 101 -303 9 111.3
IV -1 143 -143 1 116.8
1981-I +1 94 94 1 122.3
II 3 141 423 9 127.8
III 5 128 640 25 133.4
IV 7 160 1120 49 138.9
1982-I 9 125 1125 81 144.4
II 11 143 1573 121 149.4
III 13 135 1755 169 155.4
33
IV 15 187 2805 225 161.0
 Year Coded Year (X) Y XY 𝑋2 Trend
𝑌෡𝑡 = 119.56 + 2.76𝑋
Method of Least Square

0 1913 3749 1360 1913.0

෍

σ 𝑌 1913
𝑎= = = 119.56
𝑛 16
σ 𝑋𝑌 3749
𝑏= 2
= = 2.76
σ𝑋 1360

𝑌෡𝑡 = 𝑎 + 𝑏𝑥
෢𝒕 = 𝟏𝟏𝟗. 𝟓𝟔 + 𝟐. 𝟕𝟔𝑿
𝒀

34
 Analyzing Seasonal Variations
1. The Percentage of Annual Average Method
2. The Ratio to Moving Average Method
3. The Ratio to Trend Method
4. The Link Relative Method

35
 Seasonal Variations - Percentage of Annual
Average Method
• This method is used to calculate seasonal indices, which help
analyze the seasonal variation in a time series.
• It involves expressing the value of each period (e.g., quarter or
month) as a percentage of the average for the entire year.
• This method is widely used when dealing with periodic data, like
monthly sales or quarterly production.

36
 What Are Seasonal Indices?
Seasonal indices are numerical values that represent the degree to
which data points in a particular time period (e.g., a month, quarter,
or season) deviate from the average value over the year. These
indices help quantify seasonal variations in time series data and
show whether a specific time period is stronger or weaker relative to
the overall average.
• Seasonal indices > 100%: Indicate above-average performance.
• Seasonal indices < 100%: Indicate below-average performance.
• Seasonal indices = 100%: Indicate performance equal to the average.

𝒔𝒆𝒂𝒔𝒐𝒏𝒂𝒍 𝒂𝒗𝒆𝒓𝒂𝒈𝒆
𝑺𝒆𝒂𝒔𝒐𝒏𝒂𝒍 𝑰𝒏𝒅𝒆𝒙 = × 𝟏𝟎𝟎
𝒂𝒏𝒏𝒖𝒂𝒍 𝑨𝒗𝒆𝒓𝒂𝒈𝒆 37
Seasonal Variations - Percentage of Annual
Average Method- Example

Quarters 𝐀𝐧𝐧𝐮𝐚𝐥
𝐘𝐞𝐚𝐫
I II III IV Total Average
1979 72 98 79 106 355 88.75
1980 79 122 101 143 445 111.25
1981 94 141 128 160 523 130.75
1982 125 143 135 187 590 147.5

38
Seasonal Variations - Percentage of Annual
Average Method- Example

Quarters 𝐀𝐧𝐧𝐮𝐚𝐥
𝐘𝐞𝐚𝐫
I II III IV Total Average
1979 72 98 79 106 355 88.75
1980 79 122 101 143 445 111.25
1981 94 141 128 160 523 130.75
1982 125 143 135 187 590 147.5
Mean 92.5 126 110.75 149 119.56

39
 Seasonal Variations - Percentage of Annual
Average Method- Example
𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐀𝐯𝐞𝐫𝐚𝐠𝐞
𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐈𝐧𝐝𝐞𝐱 = × 𝟏𝟎𝟎
𝐀𝐧𝐧𝐮𝐚𝐥 𝐀𝐯𝐞𝐫𝐚𝐠𝐞

• 𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐈𝐧𝐝𝐞𝐱 > 𝟏𝟎𝟎% Indicates above average performance.

• 𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐈𝐧𝐝𝐞𝐱 < 𝟏𝟎𝟎% Indicates below average performance.
• 𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐈𝐧𝐝𝐞𝐱 = 𝟏𝟎𝟎% Indicates performance equal to the average.

40
Seasonal Variations - Percentage of Annual
Average Method- Example
Seasonal Annual Seasonal
Quarters
Average Average Index (%)
I 92.5 77.37
II 126 105.39
119.56
III 110.75 92.63
IV 149 124.62
Total 400.01

𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐀𝐯𝐞𝐫𝐚𝐠𝐞
𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐈𝐧𝐝𝐞𝐱 = × 𝟏𝟎𝟎
𝐀𝐧𝐧𝐮𝐚𝐥 𝐀𝐯𝐞𝐫𝐚𝐠𝐞

41
 Seasonal
Quarters Interpretation
Index (%)
Below average: This quarter performs significantly worse than the
yearly average, indicating a weak period for the activity measured
I 77.37<100
(e.g., sales, production). Efforts might be needed to understand and
address this decline.
Above average: This quarter shows a slightly stronger performance
II 105.39>100 than the average, suggesting it is a relatively stable or productive
period.
Slightly below average: This quarter is weaker but not drastically so.
III 92.63<100 It may require moderate intervention to boost performance.

Strongly above average: This is the best-performing quarter, far

exceeding the average, making it a peak period for the measured
IV 124.62>100 activity. Planning for higher demand or production during this time
might be necessary.
42
 Seasonal Variations - Ratio to Moving
Average Method
• The Ratio to Moving Average Method is a widely used technique
for analyzing seasonal variations in time series data. It removes
the trend component and calculates seasonal indices.

43
 Year & Y-Values (TCSI) 4-quarter centered moving Seasonal relatives
Quarter average (TC) (TCSI/TC)100=SI
1979-I 72 --- ---
4-Quarter centered Moving Average

II 98 --- ---
III 79 89.6 88.2
IV 106 93.5 113.4
1980-I 79 99.2 79.6
II 122 106.6 114.4
III 101 113.1 89.3
IV 143 117.4 121.3
1981-I 94 123.1 76.4
II 141 128.6 109.6
III 128 134.6 95.1
IV 160 138.8 115.3
1982-I 125 139.9 89.3
II 143 144.1 99.2
III 135 --- ---
44
IV 187 --- ---
 Seasonal Variations - Ratio to Moving
Average- Example
Quarters
𝐘𝐞𝐚𝐫 Total 𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐈𝐧𝐝𝐞𝐱 𝐒𝐈 =
𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐀𝐯𝐞𝐫𝐚𝐠𝐞
× 𝟏𝟎𝟎
I II III IV 𝐆𝐫𝐚𝐧𝐝 𝐀𝐯𝐞𝐫𝐚𝐠𝐞

1979 --- --- 88.2 113.4 𝐆𝐫𝐚𝐧𝐝 𝐀𝐯𝐞𝐫𝐚𝐠𝐞 =

𝟑𝟗𝟕.𝟐
= 𝟗𝟗. 𝟑
𝟒
1980 79.6 114.4 89.3 121.8
𝟖𝟏.𝟕𝟕
1981 76.4 109.6 85.1 115.3 𝐒𝐈 = = 𝟖𝟐. 𝟑𝟓
𝟗𝟗.𝟑
𝟏𝟎𝟕.𝟕𝟑
𝐒𝐈 = = 𝟏𝟎𝟖. 𝟒𝟗
1982 89.3 99.2 --- --- 𝟗𝟗.𝟑
𝟗𝟎.𝟖𝟕
𝐒𝐈 = = 𝟗𝟏. 𝟓𝟏
Total 245.3 323.2 272.6 350.5 𝟗𝟗.𝟑
𝟏𝟏𝟔.𝟖𝟑
𝐒𝐈 = = 𝟏𝟏𝟕. 𝟔𝟓
𝟗𝟗.𝟑
Mean 81.77 107.73 90.87 116.83 397.2
SI 82.35 108.49 91.51 117.65 400.

45
 Seasonal
Quarters Interpretation
Index (%)
This index means the performance in the first quarter is about 82.35% of the
average trend. Indicates a weaker quarter, possibly due to reduced demand
I 82.35<100 or low seasonal activity. Example: For industries like tourism, Q1 may have
less activity due to cold weather.

A 108.49% index suggests the second quarter is 8.49% above the average
trend. Indicates a stronger quarter, likely due to increased demand or
II 108.49>100
seasonal promotions. Example: Agricultural sectors may benefit from spring
harvests or favourable weather in Q2.
At 91.51%, Q3 is slightly below the trend, marking it as a neutral to weak
quarter. May represent a stabilization phase with no significant spikes or
III 91.51<100
drops in activity. Example: Retail businesses might experience steady but not
exceptional sales in late summer.
With a 117.65% index, Q4 outperforms the trend by 17.65%, marking it as
the strongest season. Reflects periods of high activity, such as holiday
IV 117.65>100 seasons, year-end production, or final quarter spending. Example: E-
commerce and retail see peak sales in Q4 due to holiday shopping.
46
 Seasonal Variations - Ratio to Trend Method
• The Ratio to Trend Method is a statistical technique used to
analyze seasonal variations in time series data.
• It involves estimating the seasonal components by removing the
trend effects, similar to the Ratio to Moving Average Method, but
with trend values calculated using a regression or other trend
estimation method.

47
Seasonal Variations - Ratio to Trend Method- Example

Quarters
𝐘𝐞𝐚𝐫
I II III IV
1979 72 98 79 106
1980 79 122 101 143
1981 94 141 128 160
1982 125 143 135 187

48
 Year Coded Year (X) Y XY 𝑋2 Trend
𝑌෡𝑡 = 119.56 + 2.76𝑋
1979-I -15 72 -1080 225 78.2
II -13 98 -1274 169 83.7
III -11 79 -869 121 89.2
Ratio to Trend Method

IV -9 106 -954 81 94.7

1980-I -7 79 -553 49 100.2
II -5 122 -610 25 105.8
III -3 101 -303 9 111.3
IV -1 143 -143 1 116.8
1981-I +1 94 94 1 122.3
II 3 141 423 9 127.8
III 5 128 640 25 133.4
IV 7 160 1120 49 138.9
1982-I 9 125 1125 81 144.4
II 11 143 1573 121 149.4
III 13 135 1755 169 155.4
49
IV 15 187 2805 225 161.0
 Year Coded Year (X) Y XY 𝑋2 Trend
𝑌෡𝑡 = 119.56 + 2.76𝑋
Ratio to Trend Method

0 1913 3749 1360 1913.0

෍

σ 𝑌 1913
𝑎= = = 119.56
𝑛 16
σ 𝑋𝑌 3749
𝑏= 2
= = 2.76
σ𝑋 1360

𝑌෡𝑡 = 𝑎 + 𝑏𝑥
෢𝒕 = 𝟏𝟏𝟗. 𝟓𝟔 + 𝟐. 𝟕𝟔𝑿
𝒀

50
 Year Y Trend Seasonal Relative
𝑌෡𝑡 = 119.56 + 2.76𝑋 𝑌Τ𝑌෡𝑡 ∗ 100
1979-I 72 78.2 92.1
II 98 83.7 117.1
III 79 89.2 88.6
Ratio to Trend Method

IV 106 94.7 11.9

1980-I 79 100.2 78.8
II 122 105.8 115.3
III 101 111.3 90.7
IV 143 116.8 122.4
1981-I 94 122.3 76.9
II 141 127.8 110.3
III 128 133.4 96.0
IV 160 138.9 115.2
1982-I 125 144.4 86.6
II 143 149.4 95.4
III 135 155.4 86.9
51
IV 187 161.0 116.1
 Seasonal Variations - Ratio to Trend-
Example
Quarters
𝐘𝐞𝐚𝐫 Total 𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐈𝐧𝐝𝐞𝐱 𝐒𝐈 =
𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐀𝐯𝐞𝐫𝐚𝐠𝐞
× 𝟏𝟎𝟎
I II III IV 𝐆𝐫𝐚𝐧𝐝 𝐀𝐯𝐞𝐫𝐚𝐠𝐞

1979 92.1 117.1 88.6 111.9 𝐆𝐫𝐚𝐧𝐝 𝐀𝐯𝐞𝐫𝐚𝐠𝐞 =

𝟒𝟎𝟎.𝟖
= 𝟏𝟎𝟎. 𝟎𝟐
𝟒
1980 78.8 115.3 90.7 122.4
𝟖𝟑.𝟔
1981 76.9 110.3 96.0 115.2 𝐒𝐈 = 𝟏𝟎𝟎.𝟎𝟐 = 𝟖𝟑. 𝟓𝟖
𝟏𝟎𝟗.𝟓𝟑
𝐒𝐈 = 𝟏𝟎𝟎.𝟎𝟐 = 𝟏𝟎𝟗. 𝟓𝟎
1982 86.6 95.4 86.9 116.1 𝟗𝟎.𝟓𝟓
𝐒𝐈 = 𝟏𝟎𝟎.𝟎𝟐 = 𝟗𝟎. 𝟓𝟑
Total 334.4 438.1 362.2 465.6 𝟏𝟏𝟔.𝟒
𝐒𝐈 = 𝟏𝟎𝟎.𝟎𝟐 = 𝟏𝟏𝟔. 𝟑𝟖
Mean 83.6 109.53 90.55 116.4 400.08
SI 83.58 109.50 90.54 116.38 400.

52
 Seasonal
Quarters Interpretation
Index (%)
This index means the performance in the first quarter is about 83.6% of the
average trend. Indicates a weaker quarter, possibly due to reduced demand
I 83.6<100 or low seasonal activity. Example: For industries like tourism, Q1 may have
less activity due to cold weather.

A 108.49% index suggests the second quarter is 9.53% above the average
trend. Indicates a stronger quarter, likely due to increased demand or
II 109.53>100
seasonal promotions. Example: Agricultural sectors may benefit from spring
harvests or favourable weather in Q2.
At 90.54%, Q3 is slightly below the trend, marking it as a neutral to weak
quarter. May represent a stabilization phase with no significant spikes or
III 90.54<100
drops in activity. Example: Retail businesses might experience steady but not
exceptional sales in late summer.
With a 116.38% index, Q4 outperforms the trend by 16.38%, marking it as
the strongest season. Reflects periods of high activity, such as holiday
IV 116.38>100 seasons, year-end production, or final quarter spending. Example: E-
commerce and retail see peak sales in Q4 due to holiday shopping.
53
 Seasonal Variations - Link Relative Method
• The Link Relative Method is used to measure seasonal variations
in a time series by focusing on how values change between
consecutive periods.
• This method is particularly useful when detecting changes in
patterns between seasons or other periodic intervals.

54
Seasonal Variations - Link Relative Method- Example

Quarters
𝐘𝐞𝐚𝐫
I II III IV
1948 71 89 106 78
1949 71 90 108 79
1950 73 91 111 81
1951 76 97 122 89

55
Seasonal Variations - Link Relative Method- Example
Quarters Quarters
𝐘𝐞𝐚𝐫 𝐘𝐞𝐚𝐫
I II III IV I II III IV
1948 71 89 106 78 1948 -- 125.4 119.1 73.6
1949 71 90 108 79 1949 91.0 126.8 120.0 73.1
1950 73 91 111 81 1950 92.4 124.7 122.0 73.0
1951 76 97 122 89 1951 93.4 127.6 125.8 73.0
𝟖𝟗
× 𝟏𝟎𝟎 = 𝟏𝟐𝟓. 𝟒
𝟕𝟏

𝟏𝟎𝟔
× 𝟏𝟎𝟎 = 𝟏𝟏𝟗. 𝟏
𝟖𝟗
⋮
𝟖𝟗
× 𝟏𝟎𝟎 = 𝟕𝟑. 𝟎
𝟏𝟐𝟐 56
Seasonal Variations - Link Relative Method- Example
• Next, we calculate the chain relatives for the four quarterly averages, setting the value of the
first quarterly average equal to 100%.
Quarters
𝐘𝐞𝐚𝐫 𝟏𝟐𝟔. 𝟏 × 𝟏𝟎𝟎
= 𝟏𝟐𝟔. 𝟏
I II III IV 𝟏𝟎𝟎
1948 -- 125.4 119.1 73.6 𝟏𝟐𝟏. 𝟕𝟑 × 𝟏𝟐𝟔. 𝟏
= 𝟏𝟓𝟑. 𝟓
1949 91.0 126.8 120.0 73.1 𝟏𝟎𝟎

1950 92.4 124.7 122.0 73.0 𝟕𝟑. 𝟐 × 𝟏𝟓𝟑. 𝟓

= 𝟏𝟏𝟐. 𝟑𝟔
𝟏𝟎𝟎
1951 93.4 127.6 125.8 73.0
Mean 92.3 126.1 121.73 73.2
Chain Relative 100 126.1 153.5 112.36

57
Seasonal Variations - Link Relative Method- Example
• Next, we calculate the chain relatives for the four quarterly averages, setting the value of the
first quarterly average equal to 100%.

Calculate the adjusted difference

Quarters 𝟏𝟏𝟐. 𝟑𝟔 − 𝟏𝟎𝟎 = 𝟏𝟐. 𝟑𝟔
𝐘𝐞𝐚𝐫
I II III IV Quarterly difference
𝟏𝟐. 𝟑𝟔
Adjusted 100 123 147.3 103.06 = 𝟑. 𝟏
𝟒
Chain Relative 𝑰𝑰 = 𝟏𝟐𝟔. 𝟏 − 𝟑. 𝟏 = 𝟏𝟐𝟑
𝑰𝑰𝑰 = 𝟏𝟓𝟑. 𝟓 − 𝟐 𝟑. 𝟏 = 𝟏𝟒𝟕. 𝟑
𝑰𝑽 = 𝟏𝟏𝟐. 𝟑𝟔 − 𝟑 𝟑. 𝟏 = 𝟏𝟎𝟑. 𝟎𝟔

58
Seasonal Variations - Link Relative Method- Example
• Next, we calculate the chain relatives for the four quarterly averages, setting the value of the
first quarterly average equal to 100%.
𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐀𝐯𝐞𝐫𝐚𝐠𝐞
𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐈𝐧𝐝𝐞𝐱 𝐒𝐈 = × 𝟏𝟎𝟎
𝐆𝐫𝐚𝐧𝐝 𝐀𝐯𝐞𝐫𝐚𝐠𝐞
Quarters
𝐘𝐞𝐚𝐫 Total 𝟒𝟕𝟑.𝟎𝟔
I II III IV 𝐆𝐫𝐚𝐧𝐝 𝐀𝐯𝐞𝐫𝐚𝐠𝐞 = = 𝟏𝟏𝟖. 𝟑𝟒
𝟒
Adjusted 100 123 147.3 103.06 473.06
𝟏𝟎𝟎
Chain Relative 𝐒𝐈 = 𝟏𝟏𝟖.𝟑𝟒 = 𝟖𝟒. 𝟓
𝟏𝟐𝟑
Seasonal 84.5 103.94 124.47 87.09 400 𝐒𝐈 = 𝟏𝟏𝟖.𝟑𝟒 = 𝟏𝟎𝟑. 𝟗𝟒
Index 𝟏𝟒𝟕.𝟑
𝐒𝐈 = 𝟏𝟏𝟖.𝟑𝟒 = 𝟏𝟐𝟒. 𝟒𝟕
𝟏𝟎𝟑.𝟎𝟔
𝐒𝐈 = 𝟏𝟏𝟖.𝟑𝟒 = 𝟖𝟕. 𝟎𝟗

59
 Seasonal
Quarters Interpretation
Index (%)
This index means the performance in the first quarter is about 84.5% of the
average trend. Indicates a weaker quarter, possibly due to reduced demand
I 84.5<100 or low seasonal activity. Example: For industries like tourism, Q1 may have
less activity due to cold weather.

A 108.49% index suggests the second quarter is 3.94% above the average
trend. Indicates a stronger quarter, likely due to increased demand or
II 103.94>100
seasonal promotions. Example: Agricultural sectors may benefit from spring
harvests or favourable weather in Q2.
With a 124.7% index, Q3 outperforms the trend by 24.47%, marking it as the
strongest season. Reflects periods of high activity, such as holiday seasons,
III 124.47<100
year-end production, or final quarter spending. Example: E-commerce and
retail see peak sales in Q3 due to holiday shopping.
This index means the performance in the first quarter is about 87.09% of the
average trend. Indicates a weaker quarter, possibly due to reduced demand
IV 87.09>100 or low seasonal activity. Example: For industries like tourism, Q4 may have
less activity due to cold weather.
60
 Cyclical Variation
• The cyclical variations can be measured by first removing the trend and
seasonal components by division and then averaging out irregular
variations.
• The simplest method of obtaining cyclical movements is called the
residual method. This method consists of removing the effects of trend,
seasonal, and irregular components from the observed time series data
in any order.
• Procedure:
• Multiply each trend value by the corresponding value of the seasonal index
to get 𝑇 × 𝑆 values.
• Eliminate trend and seasonal components by dividing each value of original
data by the corresponding 𝑇 × 𝑆 value obtained by 𝑇𝐶𝑆𝐼 ÷ 𝑇𝑆 = 𝐶𝐼 61
Year Y-values Trend Seasonal Trend × Seasonal Cyclical Irregular 3-quarter Cyclical
(TCSI) Values (T) Index TS ÷ 100 %age (CI%) moving total Relatives (C%
1979-I 72
II 98
III 79
IV 106
1980-I 79
II 122
III 101
IV 143
1981-I 94
II 141
III 128
IV 160
1982-I 125
II 143
III 135
62
IV 187
 Year Coded Year (X) Y XY 𝑋2 Trend
𝑌෡𝑡 = 119.56 + 2.76𝑋
1979-I -15 72 -1080 225 78.2
II -13 98 -1274 169 83.7
III -11 79 -869 121 89.2
IV -9 106 -954 81 94.7
1980-I -7 79 -553 49 100.2
Trend Value

II -5 122 -610 25 105.8

III -3 101 -303 9 111.3
IV -1 143 -143 1 116.8
1981-I +1 94 94 1 122.3
II 3 141 423 9 127.8
III 5 128 640 25 133.4
IV 7 160 1120 49 138.9
1982-I 9 125 1125 81 144.4
II 11 143 1573 121 149.4
III 13 135 1755 169 155.4
63
IV 15 187 2805 225 161.0
Year Y-values Trend Seasonal Trend × Seasonal Cyclical Irregular 3-quarter Cyclical
(TCSI) Values (T) Index TS ÷ 100 %age (CI%) moving total Relatives (C%
1979-I 72 78.2
II 98 83.7
III 79 89.2
IV 106 94.7
1980-I 79 100.2
II 122 105.8
III 101 111.3
IV 143 116.8
1981-I 94 122.3
II 141 127.8
III 128 133.4
IV 160 138.9
1982-I 125 144.4
II 143 149.4
III 135 155.4
64
IV 187 161.0
 Seasonal Variations - Ratio to Trend-
Example
Quarters
𝐘𝐞𝐚𝐫 Total 𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐈𝐧𝐝𝐞𝐱 𝐒𝐈 =
𝐒𝐞𝐚𝐬𝐨𝐧𝐚𝐥 𝐀𝐯𝐞𝐫𝐚𝐠𝐞
× 𝟏𝟎𝟎
I II III IV 𝐆𝐫𝐚𝐧𝐝 𝐀𝐯𝐞𝐫𝐚𝐠𝐞

1979 92.1 117.1 88.6 111.9 𝐆𝐫𝐚𝐧𝐝 𝐀𝐯𝐞𝐫𝐚𝐠𝐞 =

𝟒𝟎𝟎.𝟖
= 𝟏𝟎𝟎. 𝟎𝟐
𝟒
1980 78.8 115.3 90.7 122.4
𝟖𝟑.𝟔
1981 76.9 110.3 96.0 115.2 𝐒𝐈 = 𝟏𝟎𝟎.𝟎𝟐 = 𝟖𝟑. 𝟓𝟖
𝟏𝟎𝟗.𝟓𝟑
𝐒𝐈 = 𝟏𝟎𝟎.𝟎𝟐 = 𝟏𝟎𝟗. 𝟓𝟎
1982 86.6 95.4 86.9 116.1 𝟗𝟎.𝟓𝟓
𝐒𝐈 = 𝟏𝟎𝟎.𝟎𝟐 = 𝟗𝟎. 𝟓𝟑
Total 334.4 438.1 362.2 465.6 𝟏𝟏𝟔.𝟒
𝐒𝐈 = 𝟏𝟎𝟎.𝟎𝟐 = 𝟏𝟏𝟔. 𝟑𝟖
Mean 83.6 109.53 90.55 116.4 400.08
SI 83.58 109.50 90.54 116.38 400.

65
Year Y-values Trend Seasonal Trend × Seasonal Cyclical Irregular 3-quarter Cyclical
(TCSI) Values (T) Index(S) TS ÷ 100 %age (CI%) moving total Relatives (C%
1979-I 72 78.2 83.58 65.36 110.16
II 98 83.7 109.50 91.65 106.92 314.97 104.97
III 79 89.2 90.54 80.76 97.82 300.92 100.31
IV 106 94.7 116.38 110.21 96.18 288.33 96.11
1980-I 79 100.2 83.58 83.75 94.33 295.82 98.61
II 122 105.8 109.50 115.85 105.31 299.87 99.96
III 101 111.3 90.54 100.77 100.23 310.74 103.58
IV 143 116.8 116.38 135.93 105.20 297.39 99.13
1981-I 94 122.3 83.58 102.22 91.96 298.13 99.38
II 141 127.8 109.50 139.64 100.97 298.91 99.64
III 128 133.4 90.54 120.78 105.98 305.66 101.89
IV 160 138.9 116.38 161.65 98.98 308.53 102.87
1982-I 125 144.4 83.58 120.69 103.57 289.96 96.65
II 143 149.4 109.50 163.59 87.41 286.93 95.64
III 135 155.4 90.54 140.70 95.95 283.16 94.39
66
IV 187 161.0 116.38 187.37 99.80
 Cyclical
Interpretation
Relative (%)
• e.g., 104.97:
• This indicates an upward cyclical fluctuation.
• The observed value is 4.97% above the long-term trend.
C>100
• Example: If 104.97 is the cyclical relative for a quarter, it suggests
that the data was influenced positively by cyclical factors during
that period.
• e.g., 96.11:
• This indicates a downward cyclical fluctuation.
C<100 • The observed value is 3.89% below the long-term trend.
• Example: A cyclical relative of 96.11 for a quarter indicates
weaker performance due to negative cyclical factors.
• The cyclical effect is neutral.
C=100 • The observed value aligns perfectly with the trend.
67
 Analyzing the Irregular Variations
• Irregular variations, also known as random variations, are
unpredictable fluctuations in a time series that occur due to
random or unexpected events.
• These variations do not follow a pattern and are usually caused by
one-time events such as natural disasters, political instability,
strikes, or pandemics.
• The irregular movements of a time series are combined by
dividing the combined irregular variations by the corresponding
values of the cyclical relatives; that is
𝐶×𝐼
𝐼=
𝐶 68
Year Y-values Trend Seasonal Trend × Seasonal Cyclical Irregular 3-quarter Cyclical
(TCSI) Values (T) Index(S) TS ÷ 100 %age (CI%) moving total Relatives (C%
1979-I 72 78.2 83.58 63.36 113.64
II 98 83.7 109.50 91.65 106.92 318.38 106.13
III 79 89.2 90.54 80.76 97.82 300.92 100.31
IV 106 94.7 116.38 110.21 96.18 288.33 96.11
1980-I 79 100.2 83.58 83.75 94.33 295.82 98.61
II 122 105.8 109.50 115.85 105.31 299.87 99.96
III 101 111.3 90.54 100.77 100.23 310.74 103.58
IV 143 116.8 116.38 135.93 105.20 297.39 99.13
1981-I 94 122.3 83.58 102.22 91.96 298.13 99.38
II 141 127.8 109.50 139.64 100.97 298.91 99.64
III 128 133.4 90.54 120.78 105.98 305.66 101.89
IV 160 138.9 116.38 161.65 98.98 308.53 102.87
1982-I 125 144.4 83.58 120.69 103.57 289.96 96.65
II 143 149.4 109.50 163.59 87.41 286.93 95.64
III 135 155.4 90.54 140.70 95.95 283.16 94.39
69
IV 187 161.0 116.38 187.37 99.80
Year Y-values Cyclical Irregular %age Cyclical Relatives I
(TCSI) (CI%) (C%)
1979-I 72 113.64
II 98 106.92 106.13 1
III 79 97.82 100.31 0.98
IV 106 96.18 96.11 1
1980-I 79 94.33 98.61 0.96
II 122 105.31 99.96 1.05
III 101 100.23 103.58 0.97
IV 143 105.20 99.13 1.06
1981-I 94 91.96 99.38 0.93
II 141 100.97 99.64 1.01
III 128 105.98 101.89 1.04
IV 160 98.98 102.87 0.96
1982-I 125 103.57 96.65 1.07
II 143 87.41 95.64 0.91
III 135 95.95 94.39 1.01 70
IV 187 99.80
 I Interpretation
• e.g., 1.00
• The irregular variation is neutral.
Value=1
• There is no additional deviation from the expected values based
on the trend and systematic components.
• e.g., 0.98
• The actual value is 2% below the expected value.
Value<1
• This could indicate a minor negative anomaly, such as a slight
decline due to unforeseen circumstances.
• e.g., 1.01
• The actual value is 1% above the expected value.
Value>1
• This suggests a minor positive anomaly, such as a sudden spike in
demand or favourable unexpected factors.

71
Acknowledgment
• [Peter Andrew Bruce] Practical Statistics for Data Scientists
• [David Forsyth] Probability and Statistics for Computer Science
• [Michael Baron] Probability and Statistics for Computer Scientists
• .

72