FUNDAMENTALS OF APPLIED STATISTICS

2-2

2-1. INTRODUCTION
variable taken at successive points in time
(A set of ordered observations of a quantitative of statistical data in chronological
is Rnown as Time Series'. In other words, arrangement in terms of
of is known as Time Series1ime,
order, i.e., in accordance with occurrence time, to relate all phenomenon to
years, months, days, or hours,
is simply a device that enables one

a set of common, stable reference points.

in the field of Economic and Business
Such series have a unique important place
and production of various
Statistics since the data relating to prices, consumption
and bank clearings; sales and profits in a
commodities; money in circulation; bank deposits
national income and foreign
departmental store, agricultural and industrial production; are all
shares in a stock exchange market, etc.,
exchange reserves; prices and dividends of
A time series depicts the relationship
time series data spread over a long period of time.
between two variables, one of them being time, e.g.,
the population y) of a country in
different years (t); temperature (y,) of a place on different days (t), etc.
According to Ya-lun Chou,

"A time series may be defined as a collection readings

of belonging to different time
periods, of some economie variable or composite of variables."
Mathematically, a time series is defined by the functional relationship

=f(t)
where y is the value of the phenomenon (or variable) under consideration at time t. For

example, (i) the population (y) of a country place in different years (t), (ii) the number of
or a

births and deaths ) in different months () of theyear, (i) the sale (y) of a departmental
store in different months (t) of the year, (iv) the temperature (y) of a place on different days
(0) of the week, and so on, constitute time series. Thus, if the values of a phenomenon or
variable at times t1, t .., t are y1» Y2, , n respectively, then the series

t1, t2, t3»

1 Y2 I3 Yn
constitute a time series. Thus, a time series invariably gives a bivariate distribution, one of
the two variables being time (t) and the othèr being the value (y,) of the phenomenon at
different points of time. The values of t may be given yearly, monthly, weekly, daily or even
hourly, usually but not always at equal intervals of time
If the data are segregated by time (days, months, years, etc.) the value of the variable
under consideration changes from time to time. These fluctuations are affected not by a
single force but are due to the netefect of multiplicity of forces pulling it up and down and if
these forces were in a state of equilibrium, the series would remain constant. For example,
the retail prices of a particular commodity are influenced by a number of factors, viz., the
crop yield which further depends on weather conditions, irrigation facilities, fertilizers used,
transportation facilities, consumer demand, ete.
 ANALYSIS OF TIME SERIES
23
2-2.COMPONENTS OF TIME SERIES
The various forces at work, affecting the values of a phenomenon in a time series, can be
broadly classified into the following four categories, commonly known as the
time series some or all of which are components ofa
present (in a given time series) in varying degrees.
(a) Secular Trend or Long-term Movement.
(6) Periodic Changes or Short-term Fluctuations.
) Seasonal variations, and (i) Cyclic variations.
(c) Random or Irregular Movements.
The value of
a time series
y, at any time t is regarded as the resultant of the combined
impact of above components. In the following section we shall briefly explain them one by
one.

Remarks 1. Not all economists agree as to the classification of the components used here or as to
the manner in which they are related. Some argue that there are more than four
think that trend and cyclical movements are produced by the same set of forces. components,
and some

2. During the past two or three decades, the attempt to make forecasting more scientific has led to
the development of alternate approaches such as econometrics. Econometrics
attempts to express
economic theories in mathematical models that can be tested and verified by statistical methods. It
tries to
measure the impact of one economic variable
upon another in the hope of being able to predict
futureevents. '

3. The approach in this chapter is to present the classical statistical approach to time-series
analysis, but at the same time to point out that many other possible models exist which are based on
different assumptions and which may lead to different results.
2-2.1. Trend) By secular trend or simply trend we mean the general tendency of the data
to increase or decrease during a long period of time/This is true of most of series of Business
and Economic Statistics. For example, an upward téndency would be seen in data
pertaining
to population, agricultural production, currency in circulation etc., while.a downward
tendency will be noticed in data of births and deaths, epidemics etc., as a result of
advancement in medical sciences, better medical facilities, literacy and higher standard of
living.
1. It may be clearly noted that/trend is the general, smooth, long-term, average
tendency
It is not necessary that the increáse or decline should be in the same direction throughout
the given period. It may be possible that different tendencie_ of
increase, decrease or stability
are observed in different sections of time. However, the overall
tendency may be upward,
downward or stable. Such tendencies are the result of the forces which are, more or less,
constant for a long time or which change very gradually and continuously over a long period
of timesuch as the change in the population, tastes, habits and customs of the people in a
society and so on. They operate in an evolutionary manner and do not reflect sudden
changes. For example, the effect of population increase over a long period of time on the
expansion of various sectors like agriculture, industry, education, textiles, etc., is a
continuous but a gradual process Similarly, the growth or decline in a number of economnic
time series is the interaction of forces like advances in
production technology, large-scale
production, improved marketing management and business organization, the invention and
discovery of new natural resources and the exhaustion of the existing resources and so o n
all of which are gradual
processes
2It should not be inferred that all the series must show an upward or downward
trend.
We might come across certain series whose values fluctuate round a constant reading which
2-4 FUNDAMENTALS OF APPLIED STATISTICS

does not change with time, e.g., the series of barometric readings or the temperature of a

particular place.
3.Linear and Non-linear (Curvi-linear) Trend If the time series values plotted on graph
or less, round a straight line, then the trend exhibited by
the time series is
cluster more,
termed as Linear otherwise Non-linear (curvi-linear)In a straight line trend, the time-series
values increase or decrease more or less by a constant absolute amount, i.e., the rate of

growth (or Although,

decline) is constant. in practice, linear trend is commonly used, it is
rarely obtained in economic and business dataifIn an economic and business phenomenon,
the rate of growth or decline is not of constant nature throughout| but varies considerably in
different sectors of time. Usually, in the beginning the growth is slow, then rapid which is
further accelerated for quite sometime, after which it becomes stationary or stable for somne
period and finally retards slowly.
relative term and cannot be defined exactly. In some
4The term long period of time' is a
cases a period as small as a week may be fairly long while in some cases, a period as long as 2
years may not be enough.)For example, if the data
of agricultural production for 24 months
shows an increase it wont be termed as secular change over a period of 2 years whereas if
the count of bacterial population of a culture every five minutes, for a week shows an

increase, then we would regard it as a secular change.

2-2-2. Periodic Changes. It would be observed that in many social and economic
phenomena, apart from the growth factor in a time scries there are forces at work which
prevent the smooth flow of the series in a particular direction and tend to repeat themselves
over a period of time. These forces do not act continuously but operate in a regular spasmodic
manner. The resultant effect ofsuch forces may be classified as:
A1) Seasonal variations, and (2) Cyclic variations.
41) Seasonal Variations. These variations in a time series are due to the rhythmic
forces which operate in a regular and periodic manner over a span of less than a year, i.e.,
during a period of 12 months and have the same or almost same pattern year after year.
Thus seasonal variations in a time series will be there if the data are recorded quarterly
(every three months), monthly, weekly, daily, hourly, and so on Although in each of the
above cases, the amplitudes of the seasonal variations are differènt, all of them have the
same period viz., 1 year. Thus, in a time series data where only annual figures are given,
there are no seasonal variations.Most of economic time series are influenced by seasonal
swings, e.g, prices, production and consumption of commodities; sales and profits in a
departmental store; bank clearings and bank deposits, etc., are all affected by seasonal
variations.The seasonal variations may be attributed to the following two causes
) Those resulting from natural forces. As the name suggests, the various seasons or
weather conditions and climatic changes play an important role in seasonal movements. For
instance, thesale of umbrellas pick up very fast in rainy season; the demand for electric fans
goes up in summer season; the sale of ice and ice-cream increases very much in summer: the
sales of woollens go up in winter-all being affected by natural forces, viz., weather or
seasons. Likewise, the production of certain commodities such as sugar, rice, pulses, eggs,
etc., depends on seasons. Similarly, the prices of agricultural commodities always go down at
the time of harvest and then pick up gradually.
Ki) Those resulting from man-made conventions. These variations in a time series within
a period of 12 months are due to habits, fashions, customs and conventions of the people in
the society. For instance, the sale of jewellery and ornaments goes up in marriages; the sales
 ANALYSIS OF TIME SERIES

and 2-5
profits in departmental stores
Diwali, Dussehra go up
(Durga Pooja), Christmas,considerably during marriages, and festivals like
spasmodic manner and recur year after etc. Such
variations operate in a regular
The main year.
objective of the mea_urement of
trend and study their
effects. A study of theseasonal variations is to isolate them from the
businessmen, producers, seasonal patterns is extremely useful to
formulation of policy sales-managers,
decisions regarding etc., in planning future
operations and in
requirements, selling and advertising purchase, production, inventory control, personnel
seasonal variations, a seasonal upswingprogrammes In the absence of
any knowledge of
may be mistaken as indicator of better
conditions while a seasonal business
conditions. Thus, to understandslump may be mis-interpreted as
the behaviour of the deteriorating business
the time series data phenomenon in a time series properly,
must be adjusted for
from trend and other seasonal variations. IThis is done by isolating themn
variations (S,). This components by dividing the given time series values (y) by the
technique is called seasonal
later Seg $ 25-4]. de-seasonalisation of data and is discussed in detail
KECyclic Variations. The oscillatory
Oscillation
more than one year are termed asmovements in a time series with period of
called a 'cycle'. The cyclic fluctuations. One complete periodis
cyclic movements in a time series
Business Cycle', which are
may also be referred to as the
generally attributed to the so-called
prosperity (period of boom), recession, four-phase cycle' composed of
seven to eleven depression and recovery, and
normally lasts from
years. The upswings and
cumulative nature of the economic downswings in business
forces depend upon the
eand the (affecting the equilibrium of demand
economic and commercial series,and supply)
interaction between them. Most of the
trelating to prices, production and wages, etc., are affected e.g., series
fluctuations, though more or less regular, are not by business cycles. Cyclic
22-8. Irregular (or Random) periodic.
all the series contain another factor Component.) Apart from the regular variations, almost
called the rardom or
which are not accounted for by secular trend and irregular or residual fluctuations,
seasonal and cyclic
fluctuations are purely random, erratic, unforeseen, unpredictable and arevariations./These
non-recurring and irregular circumstances which are beyond the control of due to rumerous
human
at the same time are a
part of our system such as earthquakes, wars, floods, hand but
revolutions, epidemics, etc. These isolated or irregular but famines,
floods, revolution, political upheavals, famines, etc., are also called powerful fluctuations due to
some cases the episodic
importance of irregular fluctuations may not be significant fluctuations. In
these may be very effective and might give rise to while in others
cyclie movements.
Remark. It may be noted that
isolate irregular variations and studybecause of their absolutely random character, it is not
possible to
Only rough estimates them exclusively, nor one can forecast or estimate them
precisely.
from past experience
provisions for such abnormalities.
can be obtained and accordingly one may make some

23. ANALYSIS OF TIME SERIES m

The main problems in the analysis of time series are
() To identify the forces or components at work, the net effect of whose interaction is
exhibited by the movement of a time series, and
(iü) To isolate, study, analyse and measure them independently, i.e., by holding other
things constant
 2:6 FUNDAMENTALS OF APPLIED STATISTICS
the two models
s 1 . Mathematical Models for Time Series. The following are
commonly used for the decomposition of a time series into its components
Vi) Decomposition by Additive Hypothesis (or Additive Model). According to the
additive model, a time series can be expressed as
.. (2-1)
= T +S, +C, +Ri
where y is the time-series value at time t, T represents the trend value, S,, Ci and R
represent the seasonal, cyclic and random fluctuations at time t. Obviously, the term S, will
not appear in a series of annual data. The additive model implicitly implies that seasonal
forces (in different years), eyclical forces (in different cycles) and irregular forces (in
different long term period) operate with equal absolute effect irrespective of the trend value.
As such C; (and S) will have positive or negative values, according as whether we are in an
above normal or below normal phase of the eycle (and year) and the total of positive and
negative values for any cycle ( and any year) will be zero. R, will also have positive or
negative value and in the long-term (E R) will be zero. Occasionally, there may be a few
isolated occurrences of extreme R, of episodic nature.
The additive model assumes that all the four components of the time series operate
independently of each other so that none of these components has any effect on the
remaining three.
Vi) Decomposition by Multiplicative Hypothesis (or Multiplicative Model). On the
other hand, if we have reasons to assume that the various components in a time series
operate proportionately to the general level of the series, the traditional or classical
multiplicative model is appropriate. According to the multiplicative model,
T x S, xC,xR, . (2.2)
where S, C, and R , instead of assuming positive and negative value, are indices fluctuating
above or below unity and the geometric means of S, in a year, C, in a cycle and R, in a long-
term period are unity. In a time series with both positive end negative values, the
multiplicative model can not be applied unless the time series is translated by adding a
suitable positive value. It may be pointed out that the multiplicative decomposition of a time
series is same as the additive decomposition of logarithmic values of the original time series,
L.e.,
log y = log T + log S, + log C + 1log R,
In practice, most of the series relating to economic data conform to multiplicative model
. Remarks 1. Limitations of the Hypothesis of Decomposition of a Time Series. Hypothesis of
decomposition presupposes that the trend and periodic components are determined by separate forces
acting independently so that simple aggregation of the components could constitute the series. But in
reality, it is possible that this year's value of the series will depend to some extent on last year's value
so that trend and periodic movement will get inextricably mixed up and no meaningful separation of
them will be possible. In such a case any variations of this year may affect the whole future course of
the series and no meaningful separation of trend and periodic components will be
possible.
2. Mixed Models. In addition to the additive and multiplicative models discussed above, the
components in a time series may be combined in a large number of other ways. The different models,
defined under different assumptions will yield different results. Some of the mixed models
from different combinations of additive and multiplicative models are given below: resulting
= T, C, +S,R
y = T, + S, C,R (2-2a)
= T+ S, +CG R,
 ANALYSIS OF TIME SERIES
2-7
3. The model (2-1) or (2.2) can be used to obtain a
elimination, viz., 8ubtraction or division. measure of one or more of
multiplication model, it can be isolated fromFortheexample, if trend component (T) is the componenta by
known, then using
given time series to give
Original values
Thus, for the annual
S,xCxR,-
S,xC,xR,= T
Trend valuer
data, for which the seasonal
.. (2.26)
component S, is not there, we have
=T, x C,xR, CxR,
CxR,T (2-2c)
23-2. Uses of Time
Series The time series
analysis is of greater
but also to people working in various importance in only to
not
businessman
social
or an economist
disciplines natural,
and physical sciences. Some
1. It enables us to
of its uses are enumerated below:
study past behaviour of the phenomenon under consideration, i.e., to
the
determine the type and nature of the variations in the data.
2. The segregation and study of the various components is of
businessman in the planning of future paramount importance to a
policy decisions. operations and in the
formulation of executive and

3.
It helps to compare the actual current performance of accomplishments with
expected ones (on the basis of the past performances) and the
variations, if any. analyse the causes of such
4. It enables us to
predict or estimate or forcast the behaviour of the phenomenon in
future which is very essential for business
5. It helps us to compare the
planning.
changes in the values of different phenomenon at different
times or places, etc.
In the following sections we shall discuss various techniques for the measurement of
different components
2-4MEASUREMENT OF TREND
r Trend can be studied and/or measured by the following methods:
(i) Graphic (or Free-hand Curve Fitting) Method,
(i) Method of Semi-Averages,
(iit) Method of Curve Fitting by Principle of Least Squares, and
(iv) Method of Moving Averages.
We shall now discuss each of these methods in detail.
241. Graphic Method) A free-hand smooth curve obtained on plotting the values
against t enables us to form an idea about the general trend' of the series. Smoothing of the y
curve eliminates other components, viz. regular and
irregular fluctuations.
This method does not involve any complex mathematical techniques and can be used to
describe all types of trend, linear and non-linear. Thus, simplicity and flexibility are strong
points of this method. Its main drawbacks are:
(i) The method is very subjective, i.e., the bias of the person handling the data plays a
very important role and as such different trend curves will be obtained by different persons
for the same set of data. As such 'trend by inspection' should be attempted only by skilled
and experienced statisticians and this limits the utility and popularity ofthe metho
(i) It does not enable us to measure trend.
242. Method of Semi-averages. In this method, the whole data is divided into two
parts with respect to time, e.g, if we are given y, for t from 1991-2002, i.e., over a period of 12
years, the two equal parts will be the data from 1991 to 1996 and 1997 to 2002. In case of odd
 FUNDAMENTALS OF APPLIED STATISTICS
2-8
to the
obtained by omitting the value corresponding
number of years the two parts are middle year, viz.
the value corresponding to
middle year, e.g., for the data from 1991-2001, and plot these two
arithmetic m e a n for each part
1996 being omitted. Next we compute the covered by each part.
(means) against the mid-values of the respective time-periods and may be
averages is the required trend liñe
The line obtained on joining these two points
or future values.
extended both ways to estimate intermediate of each part would
etc. the centering of average
Remark. For even number of years like 8, 12, 16,
for period
1997-2002 (n 12), let the two averages beX1, (say)
problems, eg., from the data
=
create
1991-2002. Here will be plotted against the mean of two mid-
X1,
1991-1996 and X2 say), for the period 1991-1996, i.e., against 1st July
1993. Similarly, for the period
values, viz. 1993 and 1994 for the period
1997-2002.
obvious advantage of this method is its
Merits 1. As compared with graphic method, the
it would get the same results. Moreover,
objectivity in the sense that everyone who applies
we can also estimate the trend values.

2. It is readily comprehensible as compared to

the 'method of least squares' or the
moving average method'.
method linear relationship between the plotted points
Limitations. This assumes
mean as an average also stand
which may not exist. Moreover, the limitations of arithmetic
in its way.
Erample 2-1. Fit a trend line to the following data by the method of semi-averages
Bank Clearances (Rs. Crores) Year Bank Clearances (Rs. Crores)
Year
1992 53 1999 87
1993 79 2000 79
1994 76 2001 104
1995 66 2002 97
1996 69 2003 92
1997 94 2004 101
1998 105
Solution. Here since n =13 (odd), the two parts would consist of 1992 to 1997 and 1999
to 2004, the year 1998 being omitted.
X = Average sales for first part
120

672-83 (Rs. crores) 110 ACTUAL DATA

100
X = Average sales for second part
90
= 93-33 (Rs. crores)
80
As explained in Remarkto $ 2-4-2, 70 TREND LINE
1st
X and X2 and plotted against2001
will be 60
July 1994 1st July
50
respectively, as given in Fig. 2-1

Joining the points A [1994, X1

the trend
and B [2001, X), we get
Fig. 2.1: Trend by the Method of Semi-averages
line [Fig. 2-1].
ANALYSIS OF TIME SERIES 2-9
7

2-4-3. Method of Curve Fitting by Principle of Least Squares. The principle of

6ast squares is the most popular and widely used method of fitting mathematical functions
to a given set of datal Themethod yields very correct results if sufñiciently good appraisal of
the form of the function to the fitted is obtained either by a scrtutiny of the graphical plot of
the values over time or by a theoretical understanding of the mechanism of the variable
change.An examination of the plotted data often provides an adequate basis for deciding
upon the type of trend to use, Apart from the usual arithmetic scales, semi-logarithmic
or

doubly-logarithmic scales may be used for the graphical representation of the data. The
various types of curves that may be used to describe the given data in practice are:
dfy is the value of the variable corresponding to time t)
i) A straight line: = a+bt
(it) Second degree parabola: = a +bt + ct2
ii) kth-degree polynomial: ag+ at +agt+.. +a tR
(iv) Exponentialcurves: =a b
log y = log a +t log b =A + Bt, (say).
() Second degree curve fitted to logarithms:
Ye =a btc
logy=log a + t log b +t2 logc =A +Bt +Ct?, (say).
(ui) Growth curves:
(a) It =a +b¢ (Modified Exponential Curve)
(6) abc (Gompertz curve)
log =
=A +Bet (say)
log a +¢ logb
k
(c)
1+exp (a +bt)Logistic curve)
Remark. For deciding about the type of trend to be fitted to a given set of data, the following
points may be helpful:
) When the time series is found to be increasing or decreasing by equal absolute amounts, the
straight line trend is used. In this case, the plotting of the data will give a straight line graph.
(ii) The logarithmic straight line (exponential curve y = ab') is used as an expression of the secular
movement, when the series is inereasing or decreasing by a constant percentage rather than a constant
absolute amount. In this case, the data plotted on a semí-logarithmeic scale will give a straight line
graph.
(iii) Second degree curve fitted to logarithms may be tried for trend fitting if the data plotted on a
semi-logarithmic scale is not a straight lone graph but shows curvature, being concave either upward
or downward.
Alternatively, approximations about the type of the curve to be fitted can be made by use of the
following theorem based on finite differences
The nth differences A"y, A (log y), A* (1/y) of any general polynomial y, of nth degree in t is
constant and (n + 1)th diffërences are equal to zero."
For further guidelines, the following statistical tests
based on the calculus of finite differences
may be applied.
We know that for a polynomial y, of nth
degree t,
in
A'y, constant, r=n

0, T>n
 FUNDAMENTALS OF APPLIED STATISTIcs
2-10
interval of differencing and Ay,
where A is the difference operator given by Ay, =yi+h -Yt ,h being the
is the rth differences of y.
1. If Ay, constant, use straight line trend.
=

trend.
2. If Ay, constant, use a second degree (parabolic)
=

trend curve.
(log y,) =constant, use exponential
3. If A
fitted to logarithms.
second degree curve
4.If A (log y,) constant, use =

modified exponential, Gompertz and Logistic

curves can be
6. The growth curves, viz.,
approximated by the constancy of the ratios:
A(logy) [A(1/y)1
4 4(1/y-)S
1 1 4(logy-1)
Ay1-1
respectively, for all possible values of t,
The following tests may also be used:
constant percentage, use
modified exponential curve.
6. If Ay, tends to decrease by a or Logistic c u r v e .
use a Gompertz curve
I5 iRAy resembles a skewed frequency curve,
A y
the straight line trend
of Straight Line by Least Squares Method. Let
Fitting timet be given by the equation:
between the given time-series values y) and .. (2-3)
+bt a
the deviations
consists in minimizing the sum of squares óf
Principle of least squares we have to
estimates given by (2:3). In other words,
between the given values of y, and their
different values of t,
find a and b such that for given values of y, corresponding to n
E -2 0 - a - bt)¥
t
minima of E, for variations in a and b, we should have
is minimum. For a maxima or

OE 0 = -22 (y-a- bt)

da 2 = na + bEt 1
.. (2-4)
2t =a t+ b2 t2]'
ah0 -22ty -a-bt).
which are the normal equations for estimating a and b.
The values of yt, t , 2P are obtained from the given data and the equations (2-4) can
now be solved for a and b. With these values ofa and b, the line (2-3) gives the desired trend

line.kemark. The solution of normal equations (2-4) provides a minima of E. The proof is given below:
The necesary and sufficient condition fo a minima ofE for variations in a and b are:
0E
da2 Jadb
()0, =0. and (i)A = >0 and 0 ... (**)
0E
5 da ab2
From (2:4), we get
02S E
a2n> 0 22*>0 aa
àh ah d =
2Et
A- 22t
2 2 E2 4n-E]

n- Var(0>
Hence, the solution of the least square equations (2.4), satisfies (") and (**) and, therefore, provides
a minima of E.
ANALYSIS OF TIME SERIES 2-11

Ptting of Second Degree (Parabolic) Trend. Let the second degree parabolic trend
curve be
Y=a +bt + cr2 . (2-5)
Proceeding similarly as in the case of a straight line, the normal equations for estimating
a,b and e are given by:
2y = na + b t + c2t2
Ety, = a t +b2 2+ c23 ... (2-6)
2 y =aZe + bX12 +cZ
the summation being taken over the values of the time serieg

Fitting of Exponential Curve:

. (2.7)
logy = log a + t logb
Y =A+ Bt (say), ... (2.7a)
where Y= log y: , A = loga , B =log b. .. (2.7b)
(27a) is a straight line in t and Yand thus the normal equations for estimating A and B are
2Y = nA +BX,
.. (2-7c)
Y =AX t +BX 2 }
These equations can be solved for A and B and finally on using (27b), we get
a
antilog (A); b =antilog (B).
Second Degree Curve Fitted to Logarithms. Suppose the trend curve is
Y. = a b' c .. (2-8)
Taking logarithms of both sides, we get

logy= log a + t logb +t2 log c

Y = A+ Bt+ Ct2 ... (2-8a)
where Y, = log y ; A = loga ; B = log b and C=log c .. (2-86)
Now, (28a) is a second degree parabolic curve in Y, and t and can be fitted by the
technique already explained. We can finally obtain
a =Antilog (A); b = Antilog (B) and c= Antilog (C). | T m
With these values of a, b and e, the curve (28) becomes the best second degree curve
fitted to logarithms.
Remark. The method of curve fitting by the principle of least squares is used quite often in trend
analysis partieularly whgn one is interested in making projections for future times. Obviously, the
reliability of the estimatéd (projected) values primarily depends upon the appropriatness of the form of
the mathematical function fitted to the given data. If the function is determined on the ad-hoc basis
by
the scrutiny of the plotted values, the projections based on it may be valid for the near future while, if
the study of physical mechanism of the variable change formns the basis of the selection of function, then
there is very little likelihood that the function will change for sufficiently long period and hence in this
casepeliable long term projections can be made.

Marits and Drawbacks of Trend Fitting by the Principle of Least Squares.

Meritis.The method of least squares is the most popular and widely used method of
fitting mathematical functions to a given set of observations. It has the following advantages