7/11/2020

Time Series Analysis

Prepared by
Sheikh Yasir Arafat
Lecturer of Statistics, Department of Business Administration,
Bangladesh Army International University of Science and Technology,
Cumilla Cantonment, Cumilla.

Time Series
Time Series: A time series a set of numerical measurements on a time-dependent variable of interest
arranges over a regular interval of time.

Objectives of Time Series Analysis: The main objectives of analyzing the time series data is to get a
concrete idea about the past behavior of data so that appropriate course of action for future can be
taken. However, the objective can be pointed out as follows:
• To identify the pattern and trend, and isolate the influencing factors or effects.
• To apply the idea obtained from analyzing the pattern o time series data for future planning and
control.

Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

1
 7/11/2020

Importance of Time Series Analysis

Importance of Time Series Analysis in Business Decision-Making: Time
series analysis is of great importance to business executive. It is extremely
useful for him/her in decision-making due to the following reasons:
• This is the most popular and so far the effective method for business
forecasting.
• It helps in understanding the past behavior of economic process and in
predicting the future behavior.
• It helps in planning future operations.
• It helps in evaluating current achievement.
• It facilitates comparison of same phenomenon over two or more periods.

Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

Mathematical Models for Time-Series Analysis

Mathematical Models for Time-Series Analysis: There are two mathematical models, which used for the
decomposition of a time series into different components. These are:
• Multiplicative Model
• Additive Model
Multiplicative Model: In traditional or classical time series analysis, it is assumed that there is a
multiplicative relationship among these four components. Let 𝑌𝑡 denotes the value of a series at time t.
symbolically,
𝑌𝑡 = 𝑇𝑡 × 𝑆𝑡 × 𝐶𝑡 × 𝐼𝑡
where, 𝑇𝑡 =Trend component, 𝑆𝑡 =Seasonal component, 𝐶𝑡 =Cyclical component and 𝐼𝑡 =Irregular
component.
For example, if Tt, = 450, St = 1.4, Ct = 1.6 and It = 0.9, then, Yt = 450 × 1.4 × 1.6 × 0.9 = 907.2
Additive Model: According to this model, a time series is the sum of its four components. Symbolically,
𝑌𝑡 = 𝑇𝑡 + 𝑆𝑡 + 𝐶𝑡 + 𝐼𝑡
For example, if Tt = 250, St =45, Ct = 12 and It = -6, then Tt = 250 + 45 + 12 – 6 = 301

Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

2
 7/11/2020

Components of Time Series

Components of Time Series: The important components of a time series are
• Secular Trend
• Seasonal Trend
• Cyclical Fluctuation
• Irregular Variation

Figure: Secular trend.

Secular Trend: Many time series met in practice exhibit a tendency of either growing or reducing
fairly steadily over time. This tendency of a time series data over a long period of time is called
secular trend. Some series increase slowly, some increase fast, others decrease at varying rate, and
some remain relatively constant for long periods of time. For example, despite short-run deviations
from the trend, gross domestic product of a country, prices or productions of a farm or a country,
money supply show an upward trend over the years due to increasing population. On the other hand
population growth or death rate shows a downward trend. The time series data can be represented by
a Historigram, which exhibits its long-term trend.
Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

Secular Trend
Factors affecting trend: There are several factors that affect trend in time series
data. Some of the important factors are:
(i) Population: The ever-increasing population of a country is responsible for
increasing trends in series like prices, population, production sales etc.
(ii) Technology, Institution and culture: Downward or upward trends in some factors
are caused by technological, institutional or cultural changes.
Reasons for studying Trends: The reasons for studying trend in a time series data
are pointed below:
(i) It allows us to describe the historical pattern of time series data.
(ii) It permits us to project past pattern or trends into future.
(iii) It facilitated us to eliminate the trend component from the series in order to
obtain the de-trended series that is useful for studying other components of time
series data.

Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

3
 7/11/2020

Seasonal Trend
Seasonal Trend: Seasonal variations are like
cycles, but they occur short and repetitive
calendar periods. By seasonal variation we mean
a periodic movement that repeats itself with
remarkable similarity at a regular interval of time,
the period being no longer than one year. Hence,
seasonal component of a time series data is the
repetitive and predictable movement of
observations around the trend line during
particular time intervals of the year. In order to Figure: Seasonal Variation.
measure or to detect the seasonal component, the
data must be given in small unit of time, such as
hours, days, weeks, months or quarters. Many
business and economic time series met in practice
consists of quarterly or monthly observations.
Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

Factors affecting seasonal variations

Factors affecting seasonal variations: Seasonal variations in a series
are also caused by some factors. Some of the important factors are:
(i) Climate and weather factors: Change in the climate and weather
conditions such as humidity, heat, rainfall etc. act on different product
and industries differently and cause change in demand of them.
(ii) Social customs and Religious factors: Due to some religious festival
the sales or demand of practical product varies over the year.

Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

4
 7/11/2020

Sample Size to Estimate a Population Mean

Cyclical Fluctuation: A cycle is a wave like pattern about a
long-term trend that is usually apparent over a number a
year. The term cycle refers to the recurrent variations in time
series that usually last longer than a year. One complete
period is called a cycle. Many business and economic time
series met in practice appear to exhibit oscillatory, or
cyclical, pattern unconnected with special behavior. They are
not necessarily regular, but follow rather smooth pattern of
upswings or downswings. Examples of cycles include the
business cycle that record periods of economic recession and Figure: Phases of Business Cycle.
inflation, long term product-demand cycles and cycles in the
monetary and financial sectors. There are four well-defined
periods or phases in the business cycle, namely i) prosperity,
ii) decline, iii) depression and iv) improvement. Different
phases of business cycle are shown in following graph.

Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

Irregular variations
Irregular variations: So far we have discussed three sources of variability in a time
series. If the components of a time series are found to be trend, seasonal and
cyclical, we would expect the series very smooth and rather easily projected forward
to produce forecast. But this may not be the case practice. Because of an inevitable
presence of another component known as irregular element, induced by the
multitude of factors influencing the behavior of any actual series and whose pattern
becomes rather unpredictable on the basis of past experience.
There are two reasons for identifying the irregular components in time series data.
These are:
(i) To suggest that on occasions it may be possible to explain certain moments in the
data as due to specific causes and to simplify further analysis.
(ii) To emphasis the fact that prediction of economic conditions is always subject to
certain degree of error owing to the unpredictable erratic, influences, which may
enter.

Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

5
 7/11/2020

Measurement of Component
Measurement of Trend Component: The following are the methods used to eliminate linear trend component from a given time series data:
• Graphic or free hand curve method.
• Semi-average method.
• Least squares method.
• Moving average method.
Measurement of seasonal trend: The following methods are used to the seasonal variation in a time series data:
• Method of simple averages.
• Ratio-to-trend method.
• Ratio-to-moving average method.
• Link relatives method.
Measurement of Cyclical Component: The methods, which are used, for measuring cyclical components are as follows:
• Residual method.
• Reference cycle analysis
• Direct method
• Periodogram analysis
• Harmonic analysis

Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

Trend Projection
We know that the least square method is used to find the best straight line
relationship between two variables. We will use the same methodology to develop
the trend line for time series data. Specifically, we will be using regression analysis
to estimate the relationship between time and interested variable.
The estimated regression equation describing a straight-line relationship between an
independent variable x and a dependent variable y was written as
𝑦 = 𝑎0 + 𝑎1 𝑥
To emphasize the fact that, in forecasting, the independent variable is time, we will
use t in the above equation instead of x, we will use 𝑇𝑡 in place of 𝑦. Thus, for a
linear trend, the estimated interest variable expressed as a function of time can be
written as follows
𝑇𝑡 = 𝑎0 + 𝑎1 𝑡
where, 𝑇𝑡 = trend value of the time series in period t, 𝑎0 = intercept of the trend line,
𝑎1 = slope of the trend line and 𝑡 = time.

Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

6
 7/11/2020

Trend Projection
Formula for computing the estimated regression coefficients (𝑎0 and 𝑎1 ) is
𝑡× 𝑌𝑡
𝑡𝑌𝑡 −
𝑛
𝑎1 = 𝑡 2
and 𝑎0 = 𝑌 − 𝑎1 𝑡
𝑡 2−
𝑛
where,
𝑌𝑡 = value of the time series in period t
𝑛 = number of periods
𝑌𝑡
𝑌 = average value of the time series, that is, 𝑌 = 𝑛
𝑡
𝑡 = average value of t; that is, 𝑡 = 𝑛
.

Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

Trend Projection
Example: The sales of Zihan Foods, a small grocery chain located in Cumilla
Cantonment, for 2014 through 2019 are:
Time
Develop an equation for the Year
(𝒕)
Sales (million taka)
linear trend component of this 2014 1 7
time series data. How do we 2015 2 10
interpret the equation? What is 2016 3 9
the sales forecast for 2021? 2017 4 11
2018 5 13
2019 6 8

Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

7
 7/11/2020

Trend Projection
Solution: For developing an equation for the linear trend component of
the provided time series data, we have to estimate the slope (𝑎1 ) and
intercept (𝑎0 ) coefficients. The following table contains the required
analysis for estimation. 21 58
Year 𝒕 𝒀𝒕 𝒕𝒀𝒕 𝒕𝟐 𝑡= = 3.5; 𝑌 = = 9.67
6 6
2014 1 7 7 1
2015 2 10 20 4 21 × 58
2016 3 9 27 9 6 211 −
𝑎1 = = 0.46
2017 4 11 44 16 21 2
2018 5 13 65 25
91 − 6
2019 6 8 48 36
𝑎0 = 9.67 − 0.46 × 3.5 = 8.07
Total 21 58 211 91 Therefore, 𝑇𝑡 = 8.07 + 0.46𝑡
Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

Trend Projection
Interpretation: The sales are in millions of taka. So the value 0.46 tells
us that sales increased at a rate of 0.46 million taka per year. The value
8.07 is the estimated value of sales in the year 0. That is the estimated
for 2013, which is called the base year.

To determine the point on the line for 2020, insert the t value of 7 in the
equation. Then, 𝑇𝑡 = 8.07 + 0.46 × 7 = 11.29. Thus, on the basis of
past sales, the estimate for 2020 is 11.29 millions taka.

Sheikh Yasir Arafat, Lecturer of Statistics,DBA, BAIUST.

8
 7/11/2020

THANK YOU……