13

ANALYSIS OF TIME SERIES

MEANING AND DEFINITION
TIME-SERIES
Numerical facts which are collected at different points of time take the form
Ja time series. The unit of
time may be a year, a month, a day, an hour etc.
e time is simply device that
a enables one to relate all
ommon stable phenomena to a set
reterence points. In linear regression, two
nRe and effect relationship, one of the variables can be usedvariables have a
estimate the
ther. But in time series we use time as independent variable totoestimate some
ather dependent variable.
A set of numeric observations of the
dependent variable, measured at
gecific points in time in chronological order, usually at equal intervals in order
ndetermine the relationship of time to such variable is known as time series.
Afew definitions of time series are given below:
1. A series of successive values of a variable (or some
assumed at different points of time is called a time seriesphenomenon)
or historical
series.
2. "A set of data depending on time is called a time series."
-Kenny and Keeping
3 "A time series may be defined as collection of magnitudes belonging to
different time periods of some variable or composite of variables such
as production of steel,
per capita income, gross national income, price
of tobacco, index of industrial
production." -Ya-Lun-Chou
4"A series ofvalues over a period of time is called a time series."
-P G. Moore
8
"A time
series is a set of statistical observations arranged in chronologi-
cal order." -Morris Hamburg
6. A time series is a sequence ofvalues of the same variate corresponding
to 8uccessive points in time." -Werner Z. Hirsch
CAUSES OF VARIATIONS IN TIME SERIES DATA
perience with many examples of time series has revealed that the
observ
rations are influenced by a host of forces (or causes). Some of these forces
fecting the observations continuously, some operate after equal intervals
M
nd others are erratic. The following are the main causes of variations
time series
data :
General tendency of change in the data
i) Social customs, Festivals etc.
Gii) Seasons
iv
The four phases of business: Prosperity, Decline, Depression and
Natural
(v) Recovery
Natural calamities:
ca Earthquake, Epidemic, Flood, Drought, Famine
etc.
(vi) ete.
Folitical movements/changes, war
 SAHITYA BHAWAN PUBLICATIONS
346
COMPONENTS OF TIME SERIES
The forces affecting time series data generate certain mou
fuctuations in a time series. Such characteristic movements or fut ments on
blme series are called components of a
time series. The components
of ations d studyte
on the basis of the n n
Series may be classified into different categories
IOrces. 'There are four components of a time
series:
ations Th
1. Secular Trend or Trend d u r i n g

2. Seasonal Variations/Fluctuations
f u c t u a t

3. Cyclical Variations/Fluctuations heing&

4. Irregular (or Random) Variations/Movements

Broadly speaking, the components of a time series are first classified Althous
teru i s

threecategories: nature.

(a) Long time (or term) fuctuations Th

(b) Short time (or term) fluctuations F e a ra r

(c)Erratic or Irregular or Random fluctuations. Th

The short time fluctuations are again divided into two-(1) Seasonal
tions and (i) Cyclical Variations.
Vani (i)

The components of a time series may or may not occur at the same
time Fo
OR TREND
TREND SECULAR
The component of a time series which is responsible for its
deman
general behaviour wollen
i.e., general long term movement) over a fairly long period of time as a result d from ba
some identifiable influences is called the
secular trend or trend. It is a smoth saturd=
regular and long term tendency of a particular activity to grow or decline. The trend
may be upward as well as downward. f the series neither increases nor decreases Charac
over a long term TH
(usually a minimum of 15 to 20 years) then series is without trend
(or no-trend) or with a constant trend (See
Figure 1):
UpwardTrend

ut) | u(t)
owa rd Tren u(1)| No Trend
ww
ii

iv
Purpo
Fig. 1
It is not
necessary that the trend should be in the same«direction throughoa
the given period. The time series iii
be decreasing at various rates or may be increasing slowly or increasing aee
ormay

reverse their trend from may remain relatively constant.

growth to decline or from Some nerid TE
of time. In brief, the movements
which
decline to growth ov
n Dscil a
exhibit persistent growth or
period of time are known as secular trend. The term long period of dec18 tie
concept which is influenced by the characteristic time
rel period
The formation of of the series.
rocks is a particular example of a secular eclining

death rate is an example of downward trend; population trend. p cthvit

of upward trend. growth 18 nu
The forces which are theyfo
constant over a
sovery gradually) produce the trend. For long period (or even if theyeyvaalog
vary

example, population change, tect

progress, improvement in business organisation, better medical facilitie»
Mathematically trend may be linear or non-linear.
Purpose of Measuring Trend
There are three main
1. Knowledge of past objectives of measuring trend:
behaviour
behaviour of the factors which are
: 'Trend helps in underatandin decli
responsiblo for persiatent growtn
 ANALYSIS OF TIME SERIES
he
ation : The second objective of 347
future as a.long term forecast. measuring is to project the
th curve
udyof
ther components:
other c
The third objective of neasuring trend is to
components (sez
seasonal,
t h eo t h e r c o m

cyclical, irregular) by eliminating it.

SEASONAL VARIATIONS
nent responsible for the regular rise and
he od not more than one year is
fall in the time series
called seasonal variation.
PcCur
in
uations ocCur in regular sequence and These
h, week,
are strictly periodical; the period
a month, a a day
fraction of or even a
word 'seasonal' synonymous with the season ofday,
year,
a
is
ar
ho
the year but this
monly used to denote any type of fluctuations which are periodic in
SuTre. ln brief,
e movements tharnat are regular and periodic in nature not exceeding a
called seasonal variations.
are
he forees that are responsible for seasonal variations are
Natural factors (or climate)
conventions (Holidays, Festivals, ete.)
) Man-made
are sown and harvested at certain times every year and
For example, crops
mand for labour goes during sowing and harvesting seasons, demands of
up
increase during festivals, with draws
len clothes goes up in winter, prices
number of letters posted on
m banks are heavy on first week of a month, the
Surday is larger, etc.
laracteristics/Features of Seasonal Variations
follows:
characteristics of seasonal variations are as
The main regularly year after year
within
(i) Regularity : Seasonal variations
occur

a fixed period. about the same time and

variations occur
() Fixed Proportion : Seasonaleach year.
about the same proportion occur in both directions
variations
Increase or Decrease Seasonal
upward or downward.
can easily be forecasted.
(iv) Easy Forecast Seasonal variations
Variations
pose of Measuring Seasonal measure
seasonal variations:
to
objectives
ere are three mainbehaviour of the series
Analysis of past
Forecasting the short time fluctuations fluctuations.
variations for
measuring cyclical
mination of seasonalCYCLICAL VARIATIONS

whose period of
or patterns)
The (or swings one complete
cyclical variations
movements
latiCllatory called the
are correspond to
ation is more
r e than one year business series they
hod is called
ed a cycle. In
economic
result of economic
and booms or
depressions.

Is place a s a all economic

and business
l e and take that almost
tis anatter
n of
knowledge
common

in figure 2.
aes have four distinct phases Recovery
as
shown

TOsperity, Declin
ine,
Depressions,
Phases of Cycle

Prosperity
Normal

Decline Depression
Cyclical
m o v e m e n t s

Fig. 2
 SAHITYA BHAWAN PUBLICATIONS

348
cycle; these may be up or dowr
usually lastir they
business
They stitute the
have a duration of several years
ecurrent in nature and
o or 7 to 11 years.
In the time of prosperity, ity, producti salestwo lor
ten years
conditions are high; a period of pr
employment and other
economic
followed by decline until depression is reached,
depression is folfollowed b osperity i
reached and so on. In brief,
recovery until prosperity is
are regular
and periodic in nature but exceeding a
The movements that 10
year (any where from 2 to years) are called cyclical variations,
its phases, its nature and dum.
The causes of cyclical movements,
base upon the factual evidence of the
duration a
more or less guesses but they of business cycle activities and there areda
18no single simpleexplanation ifferent
types of cycles of varying length and size.
Cyclical variations do not occur at regular intervals and are affectod
many erratic irregular and random forces which can not be isolated
identified separately. These movements also differ in intensity or amplitude ad
each phase of movement changes gradually into the phase that follows it.
Purpose
There are three important purposes to isolating cyclical components in the
time series
i) Measures of past cyclical behaviour are valuable aids in studying te
characteristic fuctuation of a business. These measures will answe
such questions as--How sensitive is this business to general cyclial
influences? What is the typical timing, amplitude and general cyclica
pattern of the firm's production sales, inventories or raw material
prices?, etc.
i) Successful businessman plan ahead, planning requires
forecasting involves a knowledge of both typical and recent cychoal
forecasting and
behaviour. Measures of typical cycles are used in the economic
school of rhythm
forecasting, which project past cycles ahead in perou
fashion.
(ii) Cyclical measures are useful in formulating policy aimed at stabilimng
the level of business
activity.
IRREGULAR VARIATIONS
The mild or violent movements
that are of random nature called

irregular variations. are

The variations produced
by occasional forces, which may operate t once
or more than once but without
any pattern, system or
jul
irregular random or erratic variations.
or regulany alood
Earthquakes, Unusual weather, Political Strikes, Fire, Wars, Famines,"
decisions, Sudden migratio01 are
many other unforseen events are the
typical causes of random varanrend
occasional forces which produce irregular
seasonal or cyclical fluctuations. variations are not attributea w
Characteristics
6) Irregular and unpredictable: By their nature the
very irregular and unpredictable. erratic var
i) No definite pattern: Irregular variations define

pattern. do not exhibit any

ii) Short period of time: The to ver

irregular
short period of time, they tend to be influences are e conn.
c o n f i n e d

d o
t i
f m e

(iv) There is no statistical techniques forironed out over long peri l a t i n ge r r a t i "

or isolatius
influences so the estimates on the causemeasuring
rely

of erratic variations
on
sound, considered human judgement.
 ANALYSIS OF TIME SERIES 349
DECOMPOSITION OF TIME SERIES
NALYSIS
nver
To d i s c o v e r and
measure any regularities which characterise the
yhentsotaa time series and to isolate them individually is known as the
time series. In other words,
sis of a time series COnsists of discovering measuring and isolating
dysis of
Anauynts of the time series or the process of studying,
o forces that nfhuence a time series is called the interpreting and
analysis of time
t n g

analysis of the movements of variables which are dependent on time

nds of time is referred to as time series analysis. The classical approach
er

time series is in term of four distinct types of variations or separate

ADalyse a
nents that
influence a time series.
of a time series individually is after called
To study th
hstudy components
the time series.
a
mposition of
MATHEMATICAL MODELS OF TIME SERIES
of time series, the traditional or classical method is to
For an analysis
Some type of relationship among
the components of a time series. The
appose are as follows
Felationships often called "Models of Time Series"
various components of a time
1.Additive Model: Here we assume that the
additive and the values are the sum of
the four components.
mes are

mbolically
Y=T+S+C+I

where Y= Observed value in the time series

T Trend
S Seasonal Variations
C Cyclical Variations
I= Irregular Variations
that any particular value in
a
Here we assume
2.Multiplicative Model the four components. Symbolically,
Serles in the product of
Y=TxSxCxI
SERIES
OF TIME
useful in practical life.
IMPORTANCE

of time series is extremely

Une series or analysisbasis for understanding past behaviour, evaluating
Tes analysis is the future operations.
It is also used for
e C o m p l i s h m e n t s , planning series. time
paring the of different
onents
importance to a
statistician towards

Analy
aetical ot time series
is of great
data pertaining
to the field of
business and
statistical
noi ucation of
omics. the data of the
the patterns in
determine
'TOver
ime Serles analysis is usedextrapolate
to
the data into
the future. Previous
and
p e r i o d of time future activity.
to forecast same factors
Thes are studied in time series analysis
is that the
ty in a similar
economic activity
e underlyin
ing assumption
pattern of
continue
nner tO
as in the
influence the future
aa time
time series is to
past. objective in analysing
Prof.W.. Z. Hirsch
Hirsc writes, "A
valuate
main economic

changes in
omena in the hope
phenon

"g,interpret and the course of the

events."

ICipating one of thee most difficult

change in
of economic time. series analysis.
dealt with
n for thebusiness and economics
causes

ical great importance not

time eries data of
Lems in data of great
ea series
SOCiologists, Biologists
makes
bueuysis of tinme series but also for
S c i e n t i s Us,

ess and economics

 SAHITYA BHAWAN PUBLICATIONS
350
series analysis
uses of time
Inshort, following are the Solutic
Time series analysis helps in u n a
1.Analysis of past behaviour:which are responsible for the variand.
the past behaviour of the factors
2. Estimates for the future: The understanding of
of the past
past b
Tstanding
variation
beha
the past trends are extremely helpful in predictins theyiourfuturand
projecting
behaviour.
3. Forecasting: Time series study helps n forecasting and plannin.
gfuture
operations.
4. Evaluation of performance : Time series analysis helps in eval..
current accomplishments or present achievements. uating
5. Comparison : Time series analysis help in comparison among
time series. Interpreting of the variations that how they are related witvariou
other, net effect of their interaction and also with the similar changes in ot
her
time series data.
6. Estimation oftrade cycles: Time series analysis helps in the estimatin
of trade cycles on the basis cyclical fluctuations which helps thebusinessman
plans and regulate their activities. In
GRAPHS OF TIME SERIES
determ
A time series is a sequence of values corresponding to successive points d
time. Time series are graphed with time on the x-axis and the other (ii
vañable (i
under consideration on the y-axis. Such a graph is called Historigram graphs d
time series can be made in two types of scales (i) natural scale on
(i) ratio seale.
With natural scale there are two
types of graphs: (i) Absolute Historigram
here we use original values. (i) Index
or relative values. Historigram, here we use index numbers analys
Sometimes False Base Line in the vertical scale is used where the Proce
of the scales which lies between zero portn
and the smallest value of the variabie s
omitted.
To draw an absolute
x-axis and original values historigram.
Take time (years, months, etc.) aou
lines. alongy-axis. Plot the points and join them by
straug
If more than one time series
made
belongs to the same period are to bepertaining to the related phenomen nd
shown on a graph then the
is kept the same. The curves of for
each time series are shown scale oun
of lines. However, if two time series are by difrere
will have two scales, one on the left expressed in two different units
and the other on the
Illustration 1 right.
With the bhelp of the fígures
given below, prepare asuitable graph:
Trends in Sales of Units by Unit
Year Trust of India
Sales of Units
(July-June)
1964-65 (Rs. crore)
1965-66 19.1
1966-77 2.2
9.2
1967-68
1968-69 15.3 Meri
17.2
1969-70
22.8
1970-71
18.0
1971-72
15.1
1972-73
23.0
 AINALY SIS
SIS OF"
OF TIME
SERIES
351
fatton

Scale I cm= 1 Year

2
y-axis: I cm =
Rs. 5 Crore

15 AA Sales Curve

TH
Year
Fig. 3
MEASUREMENT OF LONG-TERM TREND OR SECULAR TREND
In the time series analysis, the following are the principal method of
erminingtrend
) Free-hand Curve Method
) Semi-Average Method
ii) Moving Average Method
(iv) Method of Least Squares
FREE-HAND CURVE METHOD
The term free-hand' is used to any non-mathematical curve in statistical
ursis, even when it is drawn with the aid of drafting instruments.
rocedure
6) The original data are first plotted on a graph paper by taking time on
T-axis and value on y-axis.
observed.
The direction of the points is carefully a line) is drawn through
On this observation a smooth curve (may be
the plotted points.
a trend line by the free-hand
method, an attempt should be
l e fitting
ade that the tted curve conforms to the following conditions
a straight line or a
combination of
he curve should be smooth, either
long gradual curves of the dataa
b)h e trend line or curve should be drawn through the graph
above or below the trend
are equal. They
Such a way that the areas the series and approximately
as a whole
Snould be exactly equal forlast half of the series separately and as far
qual for the first half and
as possible for each major cycle. must total the
the data above the trend line
ne vertical deviations of below the line. These deviations may be
dme as vertical deviations on the edge
of a strip ot paper, one above the
arked off cumulatively
other, for comparison. vertical deviations
of the
ne obse
observations
of the
e Sum of theshould be a s small as possible.
from
squares

the tren
Murite Demerits of Free-hand Method
Merits:
) Simple: P estimating simplest method.
trend is the sin
method of
ple: Free-hand time and labour.
This method saves is flexible
in
in the sense that if can be used
Flexible: The smooth
to represent and
linear
curve

non-linear
trend.
 352 SAHITYA BHAWAN PUBLICATIONSs
(iii) Rapid approrimations : It makes possible : rapid approxima d
trend that are relatively reliable.
iv) Non-Mathematical : This makes unnecessary any mathemati.
tions of the best curve to describe the data. atical devia,
Demerits:
6) Highly subjective : The curve depends very much on the iud
d
the drawer of the curve, hence subject to personal
No two persons will draw a similar curve.
ementst
bias, mistakes,
ete.
(i) Lack of accuracy Because of the subjective element, the cum
urve is
drawn with reasonable accuracy as compared to mathematical
(ii) Bias ofprojection for future: Free-hand method curve, being
has little value as a bias of projection for future. ctive
(iv) Estimate: This method gives only an idea of trend and not
measure.
a dof.ite
Illustration 2
Find trend the help of free-hand curve method for the data given below :
Year : 1951 52 53 54 55 56 57 58 59 60 61
Production
(in lakh ton) 15 18 16 22 19 24 20 28 22 30 26
Solution
SEMI-AVERAGE METHODD
The methods of fitting a linear trend with the
30
help of semi-average method

25
Trend

20

Original Production
15

False BaseLine

Fig. 4
are as follows
(i) The number of years is even:
The
two equal parts. 'The total of the items indata
each
of the time series are
as the
ais the
divided by the number of items to obtain part is done ands Bac
average is then centred in the period of arithmetic
time
means of the two Pmu
and plotted on the from which it has been thes
graph paper. A straight line is
drawn passing throus
points. This is the required trend line.
t h ev a l u e

ii) The number of years is odd: When the

of the middle year is omitted to divide the
number of years is oda, T h e n

follow the same procedure as in (i). time series into equal pa

To Find Mathematical Form of the
Trend Line
The straight line
passing through the two points (t1, 71) and (2 a
 ANALYSIS OF TIME SERIES
353
y-y1 y2-1
t2-ti (t -t
Thevalue
y2 is the annual increase in for a
y unit change in t.
f required, a trend value for any future year may be predicted by
lic
olying the periodic
aulaiplying increment by the number
of years into the future that
esireda n d adding the result to the best trend value listed in the series.
e r i t sa n d D e m e r i t s

Herits:

This method is simple to understand.

This is
) This
is aann objective method of measuring trend, as any one applying
this method get the same trend line.
The trend line an be extended on either side in order to obtain past
or future estimates.
Demerits
The method of semi-averages assumes a straight line relationship
between the plotted points, regardless of the fact whether such
relationship exists or not.
) There is no assurance that the influence of cycle is eliminated.
) This method has an in built limitation of arithmetic mean. This method
is not suitable in case of very low or very large extreme values.
luatration3
Determine the trend line for the following data by semi-average method:
Year :1960 61 62 63 64 65 66 67 68 69 70 72 72 73 74 75
Export
Rs.crore) 10 14 13 12 10 15 17 13 15 19 21 18 17 17 1819
(Sagar, 2006)
Solution
There are 16 years (i.e., terms) in this equation. So we divide them into two
rOups of 8 terms each, one is from 1960 to 1967 and other from 1968 to 1975.

Year Export
1960 10
1961 14
1962 13 y1 = 13
Mean 8
1963 12
Mean year, ti 1963-64 = 1963.5
1964 10
=

1965 15
1966 17
1967 13
1968 15
1969 19
1970 21 Mean,
y2- 18
1971 18 1971-72 =
1971.5
1972 Mean year, t2
=

17
1977 17
1974 18
1975 19
and (t2, y2), i.e. (1971.5
(1971.5, 18)
18) on
on the
the
points (1, ) , ie. (1963.5, 13)
aph Teis y1= 13
wll be plotted against
at the semi-average of the first half.
year 1963.5
ti= of the secona halt, y2 18 will be =

and the semi-average

against the yeat 1971.5.
 ( v)Louu
(OJO sH) Lodx
a °s 000.) s>jrS
 ANALYSIS OF TIME SERIES 357
METHOD OF MOVING AVERAGES
end with the following steps is called the method of moving
terminetrend

take the first m consecutive values of the series and calculate these
or period
a
We
to be decided by us and is called the time-inter
number:

m isa
This average is taken as the trend value for the
n the moving average.
np the middle of the period covered in the computation of average.
alright. But when the time period is even, mia
a he time is odd it is between the
mid the two observations, because the
Wheg falls way
difficulty we
with a point of the series. To avoid this
not coincide the mid-points now may
further two-utem moving averages so that the
as centring of
tinme periods. 'This process is known
UWlate

with the given

ng
averages. and include the (m + 1)th
the first observation in the series the period
pWe drop and place it at the middle of
average
ation and calculate their
N e T v a t i

1)th
+
o

values.
second to (m value and
values and include the (m 1)th
+
ering
leave the first two middle of the period covering
third to
3. Now it at the
average and place
silate their
:2)Jth values.
series.
will go on till the and of the
This process
the time series is given by
Symbolically, suppose n
Time t1, t2......., Un, .
y1, Y2, JYm, * . * * n
Value
=
placed against ta
first moving average 3
Letm =3, then
y2+34 placed against ta
second moving average

=
y3y5, placed against t4
third moving average 3
are
successive moving averages
LET m=4, then the =3F4F Y5 Ty6, and:
nd so on
- 2 +y3ty4 +y5 ys

VyT2+y3 ty4 V2 4
V1+V2 2 T *, etc.
are 2
henmoving average centred
of the
light of the
determine the period of
moving determined in the light
in
determined

oI'he timeper moving duration

averages
is which cycles
for
eriod of the Periodicity is the averagebeginning with a trough,
oticity the time series. when
decreasing again
ser T last. A cycle
cle is said to be completed
maximum point
and then
asing series
decreas
reaches to the a peak the
N86ing series when beginning
with to the
next peak.
es to the nexttrough o r the rising
series reaches
to or a n integral
and then i18 taken equal
inum point average the monthly
neral, the time period ofthegiven time series. moving However, when 7 days
tiple p e r i o d i c i t y of
the
given
then a l2
months,
4 quarters,

winarterly of daily are

g rages y figures the fluctuations

elhminate
is used to
are obtained. model
additive
In case of
averages
moving
[ue of trend of the
series.
l'from Y. thus
veonly the general
general
by
subtracting
T+S+C , trend is
eliminated
Y-T=S+C+1

In case of multiplicative
ns.
fluctuations
Y by T. Thus
S+C+D are often
Ya rxSxCx1,
e called short
trend is
time
eliminated by div
 S A H I T Y A B H A W A N P U B L I C A T I O N S

358
SxCx
trend values
not only for determining t also fin
Thus this method
method is used
1s Us i hetter than free hand curve method
Thus this
vhen (i) the t
for fitting a trend whe
It is appropriate and amnlit.ndi
regular both in period
method.
$emi-average variations are or forecastingfo and
linear, (ii) the cyclical does involve current analysis
i1) the purpose of study
Merits and Limitations
) Simple: The method of moving average 18 Simple to understan and
Merits/Advantages:

easy to adopt. 1s better than free han

(i) Objectivity: The methodis of moving average hand
because it objective.
curve method
and not subjective, it can adopt the additis
to
(ii) Flexibility: It is flexible certain new values are ad
or subtraction of
the values. That is if led
not have to be done again, but we have ta
the entire calculation will
trend values.
compute some more It irons out the cyclical fluctua.
Civ) Elimination of cyclical fluctuations:
used as tool in all parts ofthe time
tions witha uniform period and is
a

series analysis.
reduces the irregular variation
(v) Reduction of irregular variations : t
also if its time period happens to be sufficiently large.
Demerits/Limitations:
values can not be determined
(1) No 7hend values for some periods: Trend number
for some periods at the beginning and at the end. Their
increases in time period of the moving averages.
(2) Determination of the period of moving averages The choice ofthe
have
period of moving average is sometimes subjective. Sometimes we
to determine the period of moving average according to the periodicny
of the series.
(3) No complete eliminationofcyelical fluctuations: If the movingare
period is obtained by averaging the period of cyclical fluctuations, theu
reau
moving average will not completely eliminate them but will
them. few
(4) Sensitive to extreme values: Moving averages are sensitive to a
very high and low values.
Tis0n

(6) Not useful for comparison: This method is not useful for compa
of the trends.
ecasting
6) No use in forecasting: This method can not be used for forel
füture trend as moving averages assume no definite mathema
of change. trend

(7) This method (the use of simple averages) gives good results ifthe
is linear or at least approximately so. If trend is non-linear, staalues
upward them a moving average will over estimate the trena
and ifit is convex upward it will under estimate.
GRAPHICAL REPRESENTATION OF ORIGINAL VALUES
AND TREND VALUES n d e r

Time series are graphed with time on the x-axis and the variap
consideration on the y-axis. Such a graph is called Historigram'. Grapr
Beries can be made in types of scales-natural scale or ratío scale. HoW
shall discuss here only natural scale.
Sometimes false base line' in the vertical scale is used wherethei t e
Bcale which lies between zero and the smallest value of the variable 1s
 ANALYSIS OF TIME
SERIES
thehe false base line, minor 359
use

the
of
on the graph. varintionn are magnifiod and
False base line
rly visible

cessary to do so. should he used they become

utely only when it is
plotthe
the originaloriginal values:
Take yenrs
p DuPlot the graph and along x-nxis and
join them straight nd original
original values
values
linen.
g,Y-AxIS.

4the
Toplot
the trend
trend values: Plot the trend valuen on the
1o P
ts hy dotted lines; the figure so obtained
by dotted same graph and join
is the trend curve.
ofMoving Average is an add Number
eriod

Iotration 1

Findtrendi
3 yearly moving average method
from the data
Years 1960 61
62 63 64 65 66 given below
67 68 69 70 71 72 73
5 7 74
Sales ('000) 9 12 11 10 8 12 13
17 19 14 13 12 15
Dat original values and trend values on
(Viherum, 2004, 00; Ravtahankar, 2006)
graph paper.
wWution
Find three-yearly moving totals and then find
s moving average against second term and so on. Thus moving averages. Put the
there will be no moving
rage for the first value (i.e., for 1960) and for the last
value (i.e., for 1974).
Year Sales ('000) 3yearly moving Three yearly
8un
1960 moving Average
1961
57
1962 5+7+9 21 7.00
9 7+9+ 12 = 28 9.33
1963 12 9+ 12+ 11 =32 10.67
1964 11 12 +11 + 10 =33 10.00
1965 10 11+ 10+8 29 9.6
1966 8 10 +8 + 12 = 30 10.00
1967 12 8+ 12 + 13 =33 11.00
1968 13 12 +13+ 17 =422 14.00
1969 17 13 +17 + 19 =49. 16.33
1970 19 17+19 +14 50 16.67
1971 14 19+ 14+ 13 =46 15.33
1972 13 1413+ 12 =399 13.00
1973 12 13+ 12 + 15 40 13.33
1974 15
20 Sales y
18 Trend T

Scale:
X-uxis:.7cm- I year

y-uxis : c m 2 Thousand Units

-X

Ycar
Fig.8
 SAHITYA BHAWAN PUBLICATIONS

360
time series .

Illustration 2 for the following

yearly moving
average
8
9 10 11 12 13 14 15
Calculate the five 7
57 55 57 61
6
4 5
54 60
65 64 61 59
Year :55 52 49 53
(Bhopal, 2007)
Price

Solution 5 yearly Moving 5 yearly Moving

Value Sum Average, T
Year
55
52 53
263
49 54
268
53 55
273
54
279 56
60
283 57
7 57
290 58
55
295 59
9 57
302 60
10 61
65 308 62 Period
11
64 310 62
12 Tllust
61 314 63
13 Fi
14 59 Y
15 65
Illustration 3
Calculate seven yearly moving average from the following data and plot Soluti
original values and trend values on the graph paper:
Years 2 3 5 6 7 8
Values 23 26 28 32 20 12 12 10
10 11 12 14 15 16
Years 9 13
Values 13 11 14 9 3 1
12
Solution ii
Value 7 yearly Moving 7 yearly Mouing iv
Year
Y Sum Average, 1
23
2 26
28
32 153 21.8
20 140 20.0
6 12 123 17.6
7 12 108 15.4
8 10 87 12.4
9 9 81 11.6
10 13 81 11.6
11 11 78 11.1
12 14 71 10.1
13 12 63 9.0
14 9
15 3
16
 ANALYSIS OF TIME SERIES 361
araph take year on x-axis and original values and trend values
d r a w

Suitable scale (Figure 9).

Price Y Scale
-axis: .6 cm = I year
Trend T y-axis: 8 cm = 5 Units

10

36 10 12 13 14 15 16
Year
Fig. 9

is an Even Number
Average
od of Moving
atration 4 averages
yearly moving
Find trend values by the method of 4 1988 1989 1990
1991
1983 1984 1985 1986 1987
: 1981
1982
lear 320
330 340 321
285 310 320 305 310
Rs.): 290 280
Sales(000
wdaution
Alternative Procedure
second and
between
Procedure and placethem
moving s u m
Ubtain four yearly
third.
average. at a time.
Pind four yearly moving two moving
averages
the proper
value.
m) Find moving s u m taking them against
mg and place
F Control moving a v e r a g e second
PProcedure the fürst
sum
between

sum
and put
Obtain fouryearly moving and so on. sums. Place the
years moving
L982) and third (1983)
centred by
taking
two terms
i) Find moving
ng s u m third (1983)
year.
required trend
against the
by 8, obtain
centred
rst moving sum
moving
sums
below.
) Dividing these c e n t r e d table given
of the Trend
values in the last column Fouryeary
Moving Sum
(000 Rs.)
Fouryearly
Sales Y Moving Sum Centred
Year (O00 Rs.) (4)
(1) (3)
(2)
1981 290 1,165 295.0
1982 280
2,360
301.9
1983 285
1,195
2,415
1984 1,220
2,465 308.1
310 1,245
1985 320
 SAHITYA BHAWAN PUBLICATIONS
362
1986 305 1,265 2,510
1987 310 1,285 2,550 3138
1,301 2,586 318.7
1988 330
1989 340 1,311 2,612 323.2
1990 321
326.5
1991 320
EXERCISE (B)
1. Determine the trend values by three yearly moving averages
Year 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
Variable 8 12 10 13 15 12 16 17
14
[Ans.10.00, 11.67, 12.67, 13.33, 14.33, 15.00,
15.67,16.00
2. Calculate the three yearly moving averages for the following time series and plot them si
the original figures on the same graph to show its secular trend ith
Year 1 2 3 4 5 6 7 8 9 10
Annual Output: 110 104 98 105 110 120 115 110 114 126
[Ans. 104, 102.3, 104.3, 111.7, 115.0, 115.0, 113.0, 113.0,
3. Find out the value of the trend by 3-yearly moving average and plot them on a
graph paper:
Year :1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985
1986 198
Value 10 15 12 18 15 22 19 24 20 26 22 30
Jabalpur, 2007) [Ans.-, 12.3, 15, 15, 18.3, 18.7, 21.7, 21, 23.3, 22.7, 26, 25.1,4
4. Find 5-yearly moving
averages for the following time series and represent them on the same
graph along with the original data :
Year 1 3
Value 110 104 78
4

105
5 6 7
8 9 10

109 120 115 110 114 122

[Ans. -, 101.2, 103.2, 105.4, 111.8, 113.6,
116.2,-1
5. Find out the value
of
trend line with help of4-yearlytrend
moving averages and plot them on a graph paper. Dra
these values:
Year 1984 1985 1986 1987 1988 1989

Production (Lakh ton) 5 7 7 6 8

Year 1990 1991 1992 1994 1995
1993
Production (Lakh ton) 9 10 9 10 11 11

6. Obtain the information about trend

Ans. 6.38, 6.75, 7.25, 7.88, 8.62, 9.25, 2,--
9.75, L0. the
using four-yearly moving averages and reprca
onthe graph:
Year :1984 1985 1986 1987 1988 1996
1989 1990 1991 1992
Y 15 16
1993 7
17 17 16 20 21
21
18 19 20 19
Ans. 16.38, 16.75, 17.25, 17.88, 9.25, 19.25,20.12,
7. Assuming a four-yearly cycle, calculate the trend by the 18.62, 19.25, 19.25, 20 a v e r a g e sifrom

the following data relating to the method of moving av

production of tea in India:
Year 1981 1982 1983 1984 1988 1989
1990

1985 1986 1987 1988 686 619

Production (lakh kg): 464 515 518 467 502 540 557 571
9.5, 663.0, 572.6
Ans, 495.8, 503.6, 511.6, 529.5, 553.0, *
 ANALYSIS OF TIME SERIES 363
cOMPUT
UTATION OFTREND VALUES BY THE METHOD
OF LEAST SQUARES
actual values of Y and ye be the computed values of Y for a
Lety b
the
variable. X.
of+time
bx be a straight line to be fitted for trend. To find the values of
=a

sum of squares of differences of the actual and computed

Letthat the
1s
least. i.e.,
ies ofY (-yo) is least
2 y-yc) = as method of least
0 is satisfied, is known
the condition
known as the line of best fit'.
wo

The line obtained by this method aislinear b

tim series data, to find trend, the values of a and
given ime
or a normal equations.
the
ebtained by
y =Na +b Zx
xy = aZx + bEx2
which data are given.
the number of pairs for
where N is of the line. b
intercept of the line on the y-axis and b is the slope
Here a is < 0). b gives the change
growth rate (ifb>0) or decline rate (ifb
isoknown as
the value of X.
for per unit change in
the value ofY
irect Method
(1, 2, 3,..) and denote by x
into natural numbers
) Convert the years
and find Za.
values and obtain Zx
( Find the squares ofx values and obtain Zay.
the x-values with corresponding y
i) Multiply
obtain Ey.
iv) Add the values of y to and
, Exy, N) in the two normal equations
(v) Put these values (y, Ex,
solve for a and b. + bx and thena
values of a and b in
the equation. y = a
vSubstitute these
various values of x.
find trend values ye for
hort-cut Method in origin a s the
first year;
the variable x from any point of time time is taken a s origin
asure in
a r e simplified
when the mid-point
th Calculations reduces to
normal equations
a 2=0. When Zx 0, then =

y =Na a 7
and Exy b b=
method. We have to specify
The other steps in case of direct
are same as
unit of time epresented by x if needed.

erite/Demerits
Merits/advantages : mathematical
because it 1s a
This method is
objective
ective :
i) nethod
Trend forofall
determining trend. The
method of least squares gives
method of
points of time
given time periods.
the trend values for all the given method
ofleast squares is the
the
line obtained by
eline of best fit: The
line
iv) Growtof best fit. of best fit gives
the annual growth rate or declina
e
h Rate: The line
rate.
Estimation: Trend value can
he obtained for any other time period.