CURVE Frr-,-,rvG

- Chapter

s. 1. lntroduCOOn5.2. Principle of Least Squares 5.3. Fitting a Straight Line .

5
Parabola 5.5. Fitting of an Exponential Curve y = aebx 5.6. Fitting of the C.4. Fitting a SC>,,.
..
"-- Curve y = air 5.8. Fitting to the Curve xy8 = b
UNey-- a)(h ~o~D
s. 7 . egr88
U M:l' • ~Ill"
1ng Of

I5.1. INTRODUCTION
The 12eneral problem of finding equations of approximatin o c urves wh · h .
- b IC fit .
.
Cun·e fi1tnng. given data .is called
Consider n pairs of values (x 1• y 1). (x2. Y2) •··· (x,,, Y,,) of two variables x and
1
about their relationship if any. we plot the values of x and yon a suita ble scale Th)· To_get a rough ide
. . . · e point ( . a
2. .... n constitute a diagram called scatter diagram and the g, ven data (.r;. v .) i == s :\, Y;),;::: I,
1 2
bivariate. An exact relationship between the variables x a nd v i.e .. of the ~, '== / ( r)' ' :··• 11 ts said to he
. . . . . . - . .. w 111ch fits th .
sets of data 1s called curve tittmg. Generally tl ts not poss ibl e to ftnd a c urve w hich a , e given
. . I . h. b .
the e.iven potnlS. We can obtam re at1ons 1p e tween .r and _v 111 the form o r straight r p sses through aII
"' . . . . tnes, curves of
second degree. tlmd degree etc .. which may g ive the best rcprcsentat1O11 o r the biv·triate d". .b .
n,e method of least-squares can be used lo gel that re present at 1. 0 11 . It is probably' the best ISln Ul1on
10
fit;
unique curve to a gi\'en data. The other methods are : g raphic al me thod. me thod o f group averages and
method of moments. Herc we will discuss the me thod o f lc::a st s4u~1res o nly.

I5.2. PRINCIPLE OF LEAST SQUARES

Let :i set of II poinb (.r,. _v,) where;= I. 2. 3 . .... 11 represent the g iven data and .r. y are varia1es. Let
)', be the C\;~ted v::ilut· of,. •
t·orrcspondinoC,, lo .r == .r.I and w hcre as • ,.I is 1he observed value of v
corresponding, to , = .,,. In l!enaul. v. is not equal o f Y .. T he diffcn:ncc hetween the observed nnd
" • I I y
npet ted vnlue~of y is called the Error of Esrimott' o r Rt>sidual <fy a nd is J c noted by E; i.e., E;= Y; - t
Clear1Ysome of thr error~. £ w,11 · be pos1t1
. . ve and others neg ati· ve. Tl111s 11· g ·ivl:, l:'(Itnl
• weightagc to
1

II

eoch t·rrw \ \ ' t' squu~ t·uch of tht·sc nnJ for111 their sum i .<' .. 1::: = L E,2
. I • . '
·r ,ivesthebes11illln6
Now if E = 0. then all the poi nt~ lit· un till' curve. O the rwi se the 111111111,um 0 1 ~ g ·. . ,rrors jsn
• • • I • •

t•Urv t, to IIk , d· 01.1.

, TI1u /, the rnrw nf bei,I fit 1
.s th ut fur
. whi.c h the s um ol · t1·H.:• :--(.jt
• wres ol the t:
111ininw111.
342
\'( p o
,/{,rr1rJ"
~
. ts he on
'l , I of fl potfl
- - - -- - - - - - - -- -- -343
parabola (best fit).
y == a + bx+ ex2
...(i )
~ 11e se y
~el I EI == y I - I -- y I - (a + bXI + ex I 2) = y I - a - bx I - ex .2-
E2 = Yi - a - bx2- cx22, ... , E = y - a - bx - ex 2
n n n n "·
·. Jv . of
' 1·Jal',, rinc1p1e )east squares, the. best
.
values of a, b, c for whkh £ 2 + £ 2 + + £ 2
I 2 ... n
5if11 ing to the P the name of the method indicates.
,cord . iJllUJll as
A . J1lln , £2
£ = '-' ; =Y; - a - X; - ex; 1s minimum
(say) 1s 2. ..
~£ ... (ii)

or . ,·mum values of£, its partial derivatives w.r.t. a, band c must be zero.
and min
·mum
fof (Tia.XI dE dE dE
------0
da - db - de -
:. . ") artially w.r.t. a, b and c respectively :
·fferentiaung (II p
01 dE
- 2'(y;-a-b x;-ct) (-1)=0
da '--
dE
db

dE
-
de
. l'fy
s,mp 1
and re-write the equations as

L Y;= Ia+b~:X ;+cix/=n a+b LX;+cix /

Ix/ y;= a Ix/+bix /+eix/

where i varies from I to 11.
These are three equations in three unknown a, b and c. Solving them simultaneously, the best
valuesof a, band c are obtained. By putting the values of a, band c in (i) the best fitting parabola is
obtained for the given points. These three equations called Normal equations and must be committed to
memory. •
Cor. l. Best fitting line. If the points lie on a straight line, the best fitting line is given by y = a + bx
where 'a' and 'b' are given by the equations L Y;= na + b Ix1
and Ix1 y 1 = a LX; + b Ix/ which
~re known as normal equations for best fitting. Solve these equations simultaneously for a and band put
10
Y:: a+ bx if .Y is dependent variable. But if xis dependent and y is independent the line of best fitting
isx::a' +b'Y !Ill d a,' b, are given
. .
by the normal equauons ~ X; = na ' + b' ~
L, L,Y;; '"'-,X;Yi = a'"'
"'-, Y; + b'

LY,2. Solve simultaneously the normal equations for a' and b' and put in x =a'+ b'y.
pl ~or. 2· Plausible values of variables. Lei the II equations in x and y for determining the most
ausible (best) values of two variables x and y ben x + b,y = c,. 0 -;,X + h2Y= C2; ... , a,l + b,,Y = c,,. The
1
 345
., , . ancously for a and h, we ge1
1/ o s1mu 11
1 fl ''" . 1 and(") · a== o, h = 13.6
f . , ( I1) • • •
~·;' 1v1fl& f andbincqn. (1), we;gl:l
OP~1' thC values o a ~
•tilting y == I3.6 x
b~I .
OPsu
. ht line to the following data consideringy as the dtpt:ndent variable:
fitastra1g -~-=--~--;; :-~- ~
16 5,Z, l 2 3 4 5
t•9f11P 5 7 9 10 11
y
best fitting line be
1..,et the y == a + bx
11tiOl1· ... 'ii
so1
. for a and b are
quauon
·. Norillal e LY= na + b L/ .... (ii I

L xy = a I,x+bI,x2 ... ' iii)

. re given in tabular fonn below :
·ous ca1culat1on a
'fhe van xy
X
y x2
I
~ 5 5 I
I
7 14 4
2
3 9 27 9
4 10 40 16
5 11 55 25
~

L>= 1s I y =42 I, xy = 141 I .t2 = 5S

-

Putting these values in (ii) and (iii)

42 = Sa+ lSb or Sa+ !Sb = 42 ... (ir)
141 = \Sa+ 55b or !Sa+ 55b = 141 ... ( \')
Solve (iv) and (v) simultaneously for a and b
a= 3.9, b = 1.5 and put in (i)
Hence, the line of best fit is
\ y = 3.9 + {.sx I
Example 5.3. Fit a straight line to the following data :

4 6 8
4 5 6
Solution L
· et the best fitting line be
.N y = a + bx ... (i)
.. onnat equation for a and b are
 LY== na+bLx

r .\)':::: a L,X + b I,x2

The ,·arious calculation are shown below in the tabular fonn.
'•. (ii)
.r
y xy xl '•• (i;,)
1
2.4 2.4
2 1 ----
3 6.0
3 4
3.6 10.8
4 9
4 16.0
6 16
5 30.0
8 36
6 48.0
64
2,x == 24 L,Y == 24 Ixy = 113_2 Ix2 == 130 -
Putting these values in eqns. (ii) and (iii)
24 = 6a + 24b
113.2 = 24a + 130b
··· (iv)
On solving eqns. (iv) and (v) simultaneously for a and b, we get
... (11)
a= 1.976, b = 0.506
On substituting the values of a and b in eqn. (i), we get

1 y= 1.976 + o .506x 1

Example 5.4. Fit a straight line to the foHowing data regardingx as the independent variables:

Solution. Let the equation of the straight line to be fitted be

y =a+ bx
Here n = 6
X y xy x2

1 1200 l 1200
2 900 4 1800
3 600 9 1800
4 2.0 0 16 800
5 110 25 550
6 50 36 300

I>=21 LY=3060 Lx = 91
2 Lxy = 6450
 -- 347

-~ • , we get _ 60 + 21b, 6450 = 21a + 91b

0 115
0,(ftfl" att 3()60 -
I ~fl j)p:'ILI _ 1361.97, b= - 243.42
·11
,,r 110
rf11 a-
pl
~
1
f~ we gel. is ::: 1361.97 - 243.42x.
.,g, J 1,oe y
11•11' •rev
)o ~c'l111 • to the following data :
·. 3 ight 11ne ~ ---r---r---:---r- - r - - ~
· fit s str 73 69 67 65 66 67
5
fllpll 5• ' 6S 70 68 67 68 64
~jB
71 1Z 70
1 69 . f he straight line to be fitted be
uauon o t
1,et we eQ y ::: a + bx ... (i)
5011JllOl1·

1) : : na + b~:>
. are
ua1100 ... (ii )
_..,,~ eq
~01''

L xy ===aLx+b2:X2 ... (iii)

1od

~ere n~~-
x2

---- y xy
X
69 4899 5041
71 4896 4624
72
68
70 5110
\
5329
73
70 4830 4761
69
68 4556 4489
67
65 67 4355 4225
66 68 4488 4356
67 64 4288 4489

~> =546 I y =548 I xy=374~2 I i = 37314

Substituting these values in eqns. (ii) and (iii}, we get

548 = 8a + 546b
37422 = 546a + 37314b
Solving, we get
Hence, the required line of best fit is

Ly = 39.5454 + o.4242x I
 .,· t
~-.
\
,,
-', " = 0
,, -. ~, •.r, ~,·,
-',
~~~--.." t:h~t tht· lint' ~ ~"'t'S tbn m~
h the ntt' lln po int lx . y) .
~otution~L..'t th: ~-s,
fi ni u$. lin t'ar fun c tion ht.·
'" = a+ hr
Tbffi thf ll(\f fl\A f t'\lU:1tio ns :m·

'-.J •" I = ,w +I>'~ x I. ••, (i)

'
L,._,·
• I•,.I = Cl'
L,_ x I. + /,~ ' -\-.~
I
'•• (ii)
13.q113_ooos (it (1i) . liii) mny he rewrill~n as
t>x - y + a = 0
b I, x,- L Y, + ,w = 0

bI .,/ - I,x ; y, + a L x;= 0

Ehmin:mng a and b between the
se equ ati ons
X
"
LX, L •V·I 11
= 0
~ 2
- l", L X·I .V·I LrI- ... (i11)
v.nicb is the req uired best fitting
linear fun cti on for the me an
po int (x. y) .

v• = -111 'L,_ .y,.

Clearly. the lin e eqn. (iv) passes
throu gh po int (x , y ) as tw o row s of
it zero. de ter mi na nt being equ al make

5.3.1. Ch ange of Origi n

If the num ber in the given data
are lar ge, we ch an ge the ori gin
an d sca le by using substitution
x- xo
u= (h is w idt h of int erv al for
h values ofx)
and
V = y - yO
for the sake of convenience and
ease in cal cul ati on wo rk.
In case of change of ori gin ,
 349
d --1d . ·
?,,G ,quicli stant an arc ciu in number. If o ,~ the mid value' then ~··" ta'" 1:.
(f/r• . of X arc ~
' fl 1ucs x- a
1wl'' IC
1 VII
,~c 1
1
s,, ,r· II :

difference of x-valu es.

con1111on
,j11g thC f" become
v11iues o ·
/I 11~
No'" 1he _ __ I.
2
o. I. 2.... , 111 - I . 111 and so L = Lu
11 1 =Oetc.
(Ill - I), .. .. . k
., ,,,., . her of x-valucs 1s even, we ta c
,,ever. ,
·r tht: n11J11 .. - b
HO'' - _ ,, _
II- h/2

, of the middle values .

l is the me.i 11 and so
where~ · values of II become -(2m - I), .. ., - 3. - I. I. 3, ... , (2m - I)
1
. case. tie
In ,111s
, ,r' == O. etc.
~,,:::O, /..-

SOLVED EXAMPLES
I le 5 7 Fit a straight line to the foil owing data :

ExamE ;.. I t~O I 1'.s I :3 I 4~5 I 6:

Solution. The number of x values is 5. Now. take the middle value 2
as the assumed mean. then

x-a x -2
1t= - - = - - =x-2
h I

y u u2 uy
X

0 l.0 -2 4 -2.0
I l.8 -I I -1.8
2 1.3 0 0 0.0
3 4.5 I I 4.5
4 6.3 2 4 12.5

I x= 10 L,Y=16.9 L" =o L>2=,o I,uv=13.3

Let the line of best fit be
y =a +bu ... (i)

Normal equation for a and b are

... (ii)
I .v=a +bL:U
L uy = a L 11 +b L112 ... (iii)
 f
I
350- -- - -- - - -- -- -- - - -
Purting the above values in (ii) and (iii)
- 16.9 = 5a+b.0
13.3 = a .0 + b. lO
On solving (il·) and (\·) for a and b. we get
a = 3.38, b = 1.33
Hence. the ~uired straight line is
Y = a + bu = 3.38 + l.33u
Bui u = x-2
Therefore. y = 3.38 + 1.33 (x - 2)
1.e .. I y = o. 72 + l.33x l
Example 5.8. Show that the line of fit to the following data is
y = 0.1x + 11.285

X 0 5 10 15

----
20 I~
y 12 15 17 22 24

Solution. The number of x values is 6. Now,

1
2 (10 + 15) = 12.5
--
30

(x - 12.5)
Let u= 2.5 '
V = y -22

X y u V
UV

---
u2
0 12 -5 - 10 50 25
s 15 -3 -7 21 9
10 17 -1 -5 5 I
20 24 3 2 6 9
25 30 5 8 40 25

Iu=0 I,v=-12 I uv =l22 1>2 =70

Let the line of best fit be
v =a+ bu ... (IJ

The normal equations are

L:V = a+b Iu
... um
L uv = a Lu + b L u 2

Putting the above values in eqns. (ii ) and (iii) ... (iv)
- 12 = 6a + b .0 ... (v)
122 = a .0 + b.70

I
 351
we get
"or a andb,
( .#
f vlflg
501
(iV) llf1
d (V) I'

f
as -2, b = 1.74
and b in (i), we get
00 a1ues o a
. g t1te v v = -2 + l.74u
1v1fl
00 50 1.74
y _ 22= - 2 + .5 (x - 12.5)
2
G== o.1x + 11.2ss 1
or
shows the number of salesman working ,aor a certain
or fbe ,10Uowing table
,r,pl• 5.9.
~--
#'~: rear
1999 2000 2001
40
2002

56
2003
60
28 38 46
fiu"'ber :-_l.--~_.__----:-~:-:-----L-~__L_~ _J
least squares to fit a straight line trend.
etbod of
vse the Ill _ 6 (even) and h = I
1011, t-Iere n -
5011.1t u = x - (mean of two middle tenns)
~ !(~
2

X - ( 200(); 200} )

= l =2x - (2000 + 200 I) =2x _ 4001

2(1)

v= y-46
and
So that the equation y = a + bx is transformed to v = A + Bu.

Normal equations are z>= nA+B~>

and Iuv= Aiu+Biu2
Here n= 6
Table is as follows :
:r y u V IIV ,,2

1998 28 -5 - 18 90 25
1999 38 -3 -8 24 9
2000 46 -1 0 0 l
2001 40 -6 I
1 -6
2002 56 30 9
3 IO
1---
2003
- 60 5 14 70 25

---- Total
2>=0 ~>=-8 Z:uv =208 Z:u2 =70
 3 52 - - - -- -- -
Pur 1i n~ 1h~•:-c , ·.1luc in non n al l'4muions.
- - - -----
·wl' !!et
- ~= M
and 20~ = 70B
A= - l .J 3:D .1nd B = 2.97 14
Tile l.' 4ua1io n of s traight line is

,. = - 1.3333 + 2.971 1 u
Puni ng 11 = 2x - 400 1 ~m d
,. = y - 46 in (i). we get
Y - 46 = - 1.3333 + 2.971 4 (2x - 400 I) ••, (ij

[ y = 5.9428x - 11 843.9047
whic h is the re q u ired strai oht
.::, me . r
■ EXERCISE 5.1

I. F it a straight line to the le1! 1u \\ ,ng Uata taking r as a de pe ndent varia~

l: \ I I I I -
1°0
-
1
29

-
.
48
2. Fit a s trn.ie.ht li ne of regressio n o f .rand,.
.
2

.
67
3

•
8
4

for the~ da ta given be low•.

61
-
..__;
.
[Ans. v :::
· 1 + 1.9.t]

I: I , : I : I : I I: I
3 . F it a s tra ig ht l ine to the fol.low ing data regard ing .r as th e ind e pe ndent vari a ble :
[Ans. y ::: 2x]

I: \ I ~ 16~ I I I 5: I
1200 2~ 1:0
-1. Fi t a straight line to the fo llowing data conside ring .r as a d e pe nd e nt variab le :
[Ans.v= 136197 - 243.42,
1

[Ans. y = 2.0253 + 0.502t]

5. Fit a s traight li ne to the follo~i ng d ata :
X 71 68 T}, 69 67 65 66 67
\" 69 72 70 70 68 67 68 68
[Ans. y = 39.54 + 0.424x)
6. Fit a s traight line y = a + bx to the follo wing da ta by the m e tho d o f le as t sq u_ares

[Ans. y = 2.6 + 0.7.r]

'
 I~ /,~
353
/ - - : :quarcs.fil a Iincar r~ lation of thdorm p O , - - - :
d oflcas ' · cd to li ft u weight W: '' 10
th, lollo, .,.,
f
, ft''
( J11
etho
hC p
ull requ1r t
,,,· . ~ l~e p is I I 00 120
v tJS'fl~w~ere 50 70____J----t----t
1· tlil1a• W()(g) 21 25
!Ans. P = 2.275 + 0.187 Wand 30.463~ kgl
f ()(g) Wis I 50 kg.
wheo . .
ate P, k n at weekly intervals are given below. Fit a straight I' .
1il11 If ta e f · me using Dlethod
6S . t ofa ca a}culate the average rate o growth per week.
we1gh ares and C --:--i-~;-r-;--r~T~--r--::-r-
· --r---
S· ffn~east sqll 2 3 4 5 6 7 8 9 10
o I
65 70-.2 75.4 81.1 87.2 95.5 .102.2 108.4
ftge
[Ans. Y = 45.74 + 6.16 x and 6.16]

coND DEGREE PARABOLA

fffl.-...,GA SE · en set of n values m
. the equation.
.

~f . te the g1v
5ubsutu
51eP 1• W ·te the oorrna1
equations for each constant.

5tepll• rt LY=
na+bix+ ci x2;
. . .
•
;.e..
Lxy = a I x + b Ix2+ c I ,t3

I x2y = +a:Ix2+bix3+cix4
and ·
the normal equations · Itaneous equations
as s1mu · for a, b. c.
Step JII. Solve d . h . J . .
these values of a. b an c m t e equallon y = a + bx + ex-. which 1s the required
Step JV. Put
~bola of best ft t.

I SOLVED EXAMPLES
Example 5.10. Fit a second degree curve of regression of y on x to the following data :
1 2
6 11 :8 \ 2~ I
Solution. The equation of second degree parabola is given by
y = a+ bx+ cx2 ... (i)
:. Normal equations are
... (ii)

Lxy = a I x+bLx2 + c ~>J ... (iii)

L x2y = a L x2 + b ~>3+ c Lx~

... (ir)
354- -
. . I ·ulat ion :tn' shl,wn hdo w in the tabu lar
Tht' \ 3fll1U~ , .I... form :
- - -, -- ;- -, - - -

~ :c r :--
~ - ; --+ -- x 2 x
1
3 4
x
I
--

"I 11 4 8
- 16
J 18 9 27 81
➔ '"!.7 16 64 54
256
Ix= ,o Iy= 62 L·t2 = 30 Lx3 = 100 Lx4 =354 108

Putting these values in (ii), (ii,) and (iv)

22 xy ::: l 9()
62 = 4a + lOb + 30c
190 = 10a + 30b + 100c
... (\I)
644 = 30a + 100b + 354 c
On solv ing (v) and (vi), (vii) simu ltan eous ••. (\11)
ly for a, ban d c we get
... (llir)
a = 3, b = 2, c = 1
On substituting the valu es of a and b and
c in (i), we get
y=3 +2. x+x 2
Exa mpl e 5.11. Fit a seco nd degree parabola in
the follo win g dat a:

: I :o I
.r 0 3 4
y 1 17 30
Solution. The equa tion of seco nd degr ee para
bola is give n by
y =a + bx+ ex?-
:. Nor mal equa tion s are ... (I)

LY = na+ bL x+ cLx2
... (ii)

I,xy = aI, x+b L.x 2+ c Lx3 ... (iii)

I,x2 y = a Lx2 + b Lx3 + c Lx4 ... (iv)

The vari ous calc ulat ion are show n belo w in the
tabu lar form :
X y x2 x3 x4 xy x',
0 1 0 0 0 0 0
l 4 1 1 1 4 4
2 10 4 8 16 20 40
3 17 9 27 81 51 153
4 30 120 480
16 64 256 ~
L ily:: 677
LX = 10 Ly= 62 I,x2 = 30 Lx3 = 100 I,x4 = 354 I. xy = 195

OU~
r f1
7
,,,~a va10cs ,n
e
. (ii), (iii) and (iv)
62 :::
Sa + I Ob + 30c:
-- - 3Ss

... (v)
.i .g,ne 5 ::: 1Oa + 30b + I OOc
195
' 1'1111111 ... (Vi)
67 7 ::: 30a + IOOb + 354c
... (vii)
..) sirnultaneously for a, b and c we get
(vi), (vii
. (V) and a = 1.2, b = I. I, c = 1.5
d . ( ')
f a and b an c m r , we get
·
01v10S
00s values o
, tillg the y = l.2 + I.Ix + l.5x2
bS11tll taki X_
00,~ nd degree parabola to the foUowing data
e: ng - (x - S) M tht
s.tZ· fit a s~ : y -1 as the dependent.variabl
, 111plB ·able and 6 7 8 9
3 2 4 S
~.,JeOt f~l'I
8 10 11 13 10 9
-1rr_ 1 7 - L ~ J _ ~
, z 6~ - -L- --= --~ ~~ - L - e be
bolic curv
y X _ x _ 5 and Y = y - 1. Let the best fit para
5011itlol1, Given - y = a + bX + cX2 ... (I)
. ns for a' b and c are
al equauo
"
y = na + b Li X + c ~ )(2
L "
:.
')'he nor!Tl ... (ii)

L xr = a LX + b L)(2 + c Ix3 ... (ii,)

... (iv)

· u5 values of are "X "x 2 etc ., are calculated below :

Li , Li
Toe vano
- X
y XY xz x2Y xJ X4

-5 20 16 -80 -64 256

-4
- I 3 9 -9 -27 81
-3
-2 0 0 4 0 -8 16
-1 1 - l 1 1 -1 I
0 3 0 0 0 0 0
I 4 4 1 4 I I
2 6 12 4 24 8 16
3 3 9 9 27 27 81 .
4 2 8 32 64 256
t-- -_ 16

LX =O LY== 13
Ix LX 2
Y= -I rx 3=0
rx 4
= 708

-----
2
LX Y:: 55 =6o
 356- - - - - - -- --- - - - PAoa:,,
,.,.,,ur.,,"""
h 1·r t • tl1 • . , ·1IUl'' pf ' \ '. 'L , .\T l' l\.' . . and alsn II = () in (ii) · ( 111
...)
"'"'o Sr",'si
S
• Ull~ 1 u 1. c-L • • t- - ,:.... a11 I . 'c.
l (111) S
13 = 00 + /, x O+ 1· x 60 llr 9a + 60r = U ·
):i = ,1 x () + /, x 60 + c x Oor 60b = 55
/>= ll.9 17
- I = 60a + b x O+ 708c or 60a + 708c = _
1 •.. (~1)
Soh ·ing simultaneously thi: l,·). (r i) and (rii) we get ... (1,;;)
a = 3.344. b = 0.917. c = 0.285, and put in (i)
Tht' cun·e of best fi t in X and Y is
y = 3.344 + 0.9 I 7X - 0.285X2 and put X -
- X- 5, }':::: Y-
or y -7=3.344+0.917 (x -5)-0.285(x- 5)2 7
= 3.344 + 0.9 I 7 x- 4.585 - 0.285 (x3 l0
- X + 25)
y = 7 + 3.344-4.585 - 7.125 + 0.917x + 2 85
· x-0.235 2
Hence. the required parabolic curve of best fit is x·

I y = -l.366 + 3.767x- 0.285x2 I

Example 5.13. Fit a parabolic curve of regression of y on x to the following data :

; : ~ I : ~ I ::: I ~:: I ~:~ I !::

Solution. Here 11 = 7 (odd)
f8
x - 2.5
Let u= -- =2x - 5 and v=y
0.5
Resul ts in tabular form are
X y u V u2 UV u2v u3 u4

1.0 I. I -3 l. l 9 - 3.3 9.9 - 2.7 81

1.5 l .3 -2 1.3 4 - 2.6 5.2 -8 16
2.0 1.6 -I 1.6 1 - 1.6 1.6 - I I
2.5 2.0 0 2.0 0 0 0 0 0
3.0 2.7 I 2.7 I 2.7 2.7 l I
3.5 3.4 2 3.4 4 6.8 13.6 8 16
4.0 4. 1 3 4. l 9 12.3 36.9 27 81
Total LU = 0 LV =16.2 1:u2 =28 I.uv =14.3 Iu 2 v =69.9 I.u 3 =0 ru4=1 96-
Let the curve to be fitted be 11 = a + bu + cu 2 so that the normal equations are
I -L - Lu
uv = a +h Lu2+ c L,,'·
Lu2v = a Lu2 + h Lu1+ c Lu4
16 2 =?a+ 28c, 14.3 == 28b, 69 9 _ 211
• . - oQ + 196c
a = 2.07, b = 0.511 , c == 0.061

11 = 2.07 + 0.51 lu + 0.06Ju2

y = 2.07+0.511(2.x- 5) + 0.061(2x- sp
= I .04 - 0. I 93x + 0.243x2.

[ y = 1.04 - O. l 93x + 0.243x2]

nd degree parabola to the following data by least square method:

:. 14, Fit a seco
~,~P~ 5, 9 1930 1931 1932 1933 1934 1935 1936
1931
m 352 356 357 358 360 361 361 360 359

Y b of x values is 9. Now take the middle value 1933 as the assum ed mean. then
.10n The num er
501ut • x- a y- b
u= and v = -
h h
x - 1933
u= and V =y- 357
i.e., 1
u = x - 1933 and V =y-357
. calculations are shown in the tabular form :
or vanous
The
,----
u y u2 u2v
;r V UV
u3 I
'
1929 -4 352 -5
I u4
20 16 -80 -~ 256
I1930 -3 356 - I 3 9 -9 - 27 81
1931 -2 357 0 0 4 0 -8 6
I1932 -l 358 I - 1 I l - I I
/1933 0 360 3 0 0 0 0 0
1934 l 36 ] 4 4 I 4 I I
J 1935 2 36] 4 8 4 16 8 16
1936 3 360 3 9
1937 9 27 27 8
4 356
I ----- 2 8 16 32 64 256
Total
tu== o
let ----- - I>== I I Iuv == 51 L:U2=60 I112,. = - 9 Iu-' = o Ll= 108
the degree e . .
quatton of second degree parabola be
... (i)
v == a + bu + cu 2
358- - - - - - ------ ------
'ilU'tti,1n~ :If\'
The ,wrnta1t"'1 •

L''= 11<1 +b L CL II+ 112

L "'' = a L" + b L u 2 + c Li/'

3nd
L 112,. = a L u 2 + b L u 3 + c L u4
'•, (i;,)
Pulling the rnlues in eqns. (ii). (iii) and (iv)
I I = 9a + 60c
51 = 60b
- 9 = 60a + 708c '•, (\I)
On sohi ng eqns. (11). (vi) and (vii ) for a, b and c we get '•, (1,~

a= 3, b = 0.85 and c == _ 0 _ •,, (11;;)

27
On soh-ing the above values in eqns. (ii), (iii) and (iv), we get
v = 3 + 0.85u - 0 .27u 2

or y - 357 = 3 + 0.85 (x - 1933) - 0.27 (x _ )2

1933
I~y
2
- 0.27x + 1044.67x - 1010135.0 8]

■ EXERCISE 5.2
1. By the method of least square. fit a parabola y == a + bx + cx2 in the f
o 11owing data :
~
X -I 0 0 1
2 0 2
[Ans. Y ::: 0.5 + I.S.~]
2. Fit a second degree parabola to the following data :
1 4 5
·~ I 3 I 3~6 1 13 6~4 I 12 62 20.86
6
31.53
[Ans. y = 4 .9 - 3.1 lx + l.257x2]
3. Fit a parabola to the following observations :
X

)'
\ -2 -1 I
\ -3. 1s : -1.39
I
2.88 5.378 1
[Ans. y = 0.621 + 2.123x + 0.123il
4. Fit a second degree parabola to the following data taking x as

2
I: 6
3
7
4
8
5
10
6
11
r - 0.27.tl
[Hint : Put X = x - 5] [Ans. y = - 1 + 3·35'
 ~ . stance R to the motion of a train at a ~peed V. Ir the law c
-- 359
~ ,tJG ·yes rest . .
, c below gt R _ a + bv2, find by the method of least square the most plausib onnectrng R
./~ ble f rm - • l e va1ue~
C ,(11c \ll . of the o
, I y 1S
,. ~od d b- 30 40
of ti all I0
20 50
10 15 21 30
V 8 [Ans. a= 6.7, h = 0.0092]
-a+ bx
+ cx2 to
the following observations :
arabo\a y - --2- -.-- -:3:-----r----:-
4---r- --=-
5 --,
fil ap I
6·
28 33 39 46
y - -3 , v=y- 33
of' ke u - x
Jliot: 1a [Ans. v = - 0.08 6 + 5.3u + 0.643u 2 or y = 22.801 + 1.442x + 0.643x2]

. 0 f the form R = a + bV + cV2 to the following data, where vis the

velocity in
,1 a relation . h train resistance in kg/quintal :
f
7, t and R ts t e
kinfhr ·?·-~
0 _,.. -4-0- r -6-0--- -i-8-0- , -l-OO- r--
l2-0-,
V -
5.5 9.1 14.9 22.8 33.3 46.0
.
Es111nate
Rwhen V= 90. [Ans. R = - I.5726 + 0.3290 v + o.oooosv2, 28.69]
. h ast square , fit of the form Y = +
8 Find t e 1_e _ _- - , ---=-----,-- ~--,-0
--=---
a ai
to the following data:

.~I
Q_ \
-I
2
I
1
O
5
I
\
I
3
I I
-
2

0 - [Ans. y = 4.1667 - 1.111 Lt2]

fss. FITTING OF AN EXPONENTIAL CURVE y =a~x

Step I. Taking logrithms on both sides, we get
1og 10 y = log 10 a+ bx log e
10
i.e. Y=A+ Bx where Y = log y, A= log a and 8 = b log c
10 10 10
Step II. Write the normal equations
i.e. IY = nA + BI.x; I.xY = A I.x + BI.x2
Step III. Solve the normal equations for A and B. Then find out a and b from
A= log toa and
B = b log 10 c.
Step IV. Put these values in the equation y = aehX, which is the required curve
of best fit.
~ FITTING OF THE CURVE y = a/'
Step I. Taking logarithms on both sides, we get,
. log 10 y= log 10 a+blog 10 x
1.e. Y=A+ bX where Y= log y, A= log a and X = log x
10 10 10
~tep II. Write the normal equations
i.e. I.Y = nA + bI.X; ~y =ALY+ br.Y2
r 1 ~tep m. Solve the normal equation as simultaneous equations for A and 8
eation A_ 1 then find out a from the
- og1oa
Slep IV. Put these values i.e. a and bin the equation y = a:ih which is the require
r! curve of best fit.

~
 360- - -
~---------:--------~ Ae1ur y ~
TTING Of THE CURVE y = ab" 4rvo Sr'\~ ,
Fl I -·d~ .,
.· ,, !(v•:irithlll~ on bot 1 s1 t: S. we gel
.,.,C,"
Step J. Takt llt- . t-
log 10 y = lug 10 a+ x log 10 b

i. e.
r = A + Bx
when:' Y = log ·v. A = log a, B = log b
Write the normal equations
S~p II . .
i.e. LY= llA + B"f.x
and IxY = ALt + B"f.x2
Step 111. Solve the normal equation as simultaneous equations for A
. . . and B th
from rhe relauons . en finct ou
. t a
a= Antilog (A) and b = Antilog (B) and b
Step rv. Put there values I. e. a and bin the equation y = air which is th .
. . e required f
o best fit c
fS.S. FITTING O THE CU~VE xy° = b Urve

Step I. Taking logarithms on both sides, we get

log 10 x +alog 10 y = log 10 b

1 I
or log 10 y = ~log 10 b- a log 10 x

1 1
i. e. Y=A+BXwhere Y=log 10 y, A= - log 10 b , B= - - andX=loo x
a · 0 a 10

As the problem reduces to Y = A + BX. hence the rest of the procedure is sirnilar as above.

■ SOLVED EXAMPLES
E~ample 5.15. Fit a curve of the best fit of the type y = aebx to the following data by the
method of least squares:

X I 7 9

y 10 12 15
Solution. To fit a curve of the form
y = aebx ...(i)
by the method of least squares, we use the following normal equations:
...(ii)
X.Y =- nA + BI.x
...(iii)
and IxY = AI.x + BI.r2
where Y = log 10 y , A= log 1~ a B = log 10 e

,·/
•
~
I, /piff//J..,
. re shown in the tabular form
ca\c\Jlal!OOS a
- 361
.:::

f r.f ·oLls Y Y =-log10Y 2 .

{· vlll'1 . . X
xY----- - ----
r-~--:.- ioooo T-~I- -.:._-t-~
'

111~ 1 11to- 1 ~J.OOOC.

I
15 1.1761 25
5.8805·
12 1.0792 49 .
7 7.5544 .
15 1.1761 81
9 10.5849
21 1.3222. 144
12 I5.8664
ly = 5._7536 Lt2 = 300
r.x:::: 34 --:-::: ...:-.J.... ..__--- --L_ ~:_eLy,=
_~40.8862
~_J
. •n (ii) and (w)
aloe 1
•11(1 wese v 5 7536 = SA + 348·
piiltl" .
40.8862 = 34A + 300B . ...(ii')
•..(1•)
. and (v) simultaneously for A-and B, we get ·
solving (iv) A= 0.9766 and B = 0.02561
011
a= Anti log (A)= ·9.4754

b = log.N· e =_· .0.059

. ·
10
and
. the values of a and b in (i), we get
On solvtng y = 9.4754 e0.059x

Example S.16. Fit a relation of the form Y = air fo_r the following data by t_be method of le~st
squares:
X 2 3 ·, 4 s 6
..
y 8.3 15.4 33.1 65.2 126.4
Solution. To fit curve of the form
y = abX ...(i)
by the method of least squares, we use the fol.lowing -nor~ial equations:
lY= nA+ BI.x ...(ii)
I.x Y = AI.x + BLi:2
where A= log 10 a, B = log 10 band Y= log!Oy
The various calculations are shown in the tabular form:

X y xY
Y =::log10 y x2
2 8.3 0.9191 4 1.8382
.3 ' 3.5416
15.4 l.1872 9
4 33.l 1.5198 16 6.0792
5 65.2 25 9.071 0
1.8142
'
6 36 12.6312
I--_ 126.4 2. 1052
4:::20
lY 7.5455 Lt-2 = 90 I.xY= 33.1812 I
L>;,,
 .>b.:_
... ,,, - , 1'1 .-, 1n11,) .rn,l{ ,;)
1 tl•' I , ' ' •
. .. .- - , .J " ·' - 5 \ + 2llH

l )1 , ,,' , 11~ 1 1, Jn.i ,,' :-1mult.rnl',,11:-I~ for A and B. we get

A = 0 .31 and B = 0.3
11 = Anti log (A )= 2.04
.rnJ h = Antil og (B) = 1.995

i y = 2.04 ( 1.995y j
ExampJe 5.1 7. BY using the meth od of least squa re find
· a relat·
tht> data: ' ion of th ~
e •0 rlll y :::: a.tb
2 that fits

---::--XJ _
Solution . To fi t a curve of the form
__.__---:-
27_._8 __.___6_: _.1__,___1_: _0 --L\__ 1:_l_j
,. = ru.b
b~ the metbcxi of least sequ ares. we use the follo wing
norm al equa tions
••,(i)
lY= 11A + bl½.
and LXY = AL½ + blx.2
.. .(ii)
where X = log 10 x ,Y = log 10 y, A= log a ... (iii)
10
The , ariou s calcu lation s are show n in the tabul ar form
:
.r y X =log 10 x Y = log10 y xi
.., XY
27.8 0.301 0 1.440 0.090 6 0.4346
3 62. 1 0.477 1 1.793 1 0.227 6 0.8555
4 11.0 0.602 1 2.041 4 0.362 5
I 5
1.229 1
161 0.699 0
I 2.206 8 0.488 6 1.5426

I I LX =2.079 2 l:Y= 7.485 3 l:X2 = 1.169 3 EXY = 4.061 8

Putti ng these value s in (ii) and ( iii)

7.485 3 = 4A + 2.079 2b ...(iv)
4 .06 18 = 2.0792A + 1.169 3b ...(v)
On so lving ( i v) and ( v) simu ltane ously for A and b, we
get
A = 0.867 8 and b = 1.93 11
a = Antil og A= 7.376
On subst itutin g the value s of a and h in (i), we get

y = 7 .376xl.93l l I
 3&3
or the gas corr~~po ndin~ to \ a ri,
pressure ,u-, "o1umt \ , llH:a u rLd
ffJC
,rJO
8
f,r' , 1~jflg
· dlilui: -SO
l
' , rf1r roll"
,,,e ' 60 70 \ 90 I OQ
e~~,,,e' y(cf11 3) 51.3 40.5 25.9 78
,11 -Z)
,i' cffl
f (1'g . 0
pV" == C.
u fltlO
ttJC eq
tJJ to h ve
e d9 e we a pvY == C •. ( I )
fil I~ 10'1• f-Je( ' p == CV-Y
,olLlt log p == log C - y log V
Y == A+ BX
Y == Iog 10 P, A= log 10 C, B =-y, X = log V
~
~ ations are
,~ere .-JllaJ equ ~y = nA + B"I.X ...(ii}
\: 'f)1e riO ALX + Br.x.2 rxr = ... (iii)
·· e shown in the tabular form:
aod . s calcU
lations _:a:_:
r :__=-~,-~:-----:----:::-T---=----
y l p
filevaflou X == log V = og XY xi
p
l' 1.69897 1.81090 3.07666 2.88650
1.77815 1.71012 3.04085 3.161 82
51.3 1.60746 2.96592
60 1.84510 3.~39
40.5 1.41330
70 1.95424 2.76193 3.81903
25,9
90 2 1.89209 7.78418 ➔
78
LX = 9 .27646 ~Y = 8.43387 LXY= 15.62954 tx2 = 17.27 176

. these values in (ii ) and (iii)

ru1ung
8.43387 = 5A + 9.27646B ... (ir)

15.62954 = 9 .27646A + 17.27176B ... (\')

Onsolving (iv) and ( v) simultaneously for A and B, we get
A= 2.22476 and B = -0.28997
y = -B = 0 .28997
C = Antilog (A)= 167.78765
On substituting the values of y and C in (i) , we get

I PV 0 28997
· = 167.78765

I EXERCISE 5 . 7
-
I.
-- X 0.0 0 .5 1.0 1.5 2.0 2.5

--- y 0.10
- 0.45 2 .15 9. 15 40.35 180.75 [Ans. y = 0.10 l 9e29963x]
 364 -
(1 8
--
10
-- -- P~o B4e ,lit'r
2. - ---
I1 .0~.i 30.1'.!H 4~o s
.ion 8 1.89 7 222. 62 7'41,~
''t-

·" \'
T
I
1.65
....
_.__
2~7 f 3
4.5
I
4
7.35
I
l
4

j
4. \ 0 2 6 8

'
150 63
I 28
I 12
I 5 .6

5. \
I 1 2 3 4 5 6
,. 1.6 4.5 13.8 40.2 125 300
-
Fit an exponentiaJ curve of the form y = air to the
follo wi ng data : rAns . y -
- 0,5580 I n,
e ·"llll
6. X

,.
1 2 3 4 5 6 7 - 1
I 87 97 11 3 129 202 195 193
-
-
[Ans. Y == 73.9656
7. X 2 3 4 5 6 (l.1681)1
I
,. 144 172.8 207.4 24 8.8 298. 5

8.
;
X

\' I 1.0
I
2
1.2
I
3
1.8
I
4
2.5
5
3.6
6 7 r~~I 86(\2~
4 .7
I 6 .6

rAns . y == 0.6824 (l .3828Y]

X 0 2 3 4 5
9. '. 7
)'
I 10
I 21
I 35
I 59 92 200
I 4~ I 6 10
I
[An s. y = I0.499 (l.7959)x]
By the method of least squares, find a relation of
the form y =axh that fits data:
10. X I 2 3 4 5
y 7.1 27.8 62.1 11 0 16 1
i.952]
[Ans . y = 7.173x

2 \ 3 4 5
2 I 4.5 8 12.5
012l9'J l
77
[Ans . Y = O. 5
 - 365

( # ~ __,..- -3 4 5 l 6 I
CUµv€ ~--1f-----:--:--1-:-:---1
4 .26 5.21 6. 1 6.8
t--
7.5
tZ• . - ~-__,
!Ans. 2.9nJ>~ 1~1 ]
d the volume of a gas are related by the equatio n
ssure an . . .
,rt,e pre- d k being constants. Fit this equatio n to the following set of observatio m.:
13, p\!"{::: k, y an. -----r--:-::----r----::--:--r--::-:--r~~--i-- - -
2) 0 .5 1.0 1.5 2.0 2.5 I 3.0
p (l<g/crn
,~ -+-----t--:-~1---;::::-::--r-~-;;:-1--::-::::----, --:-:-:-1
1.62 1.0 0.75 0.62 0.52 0.46
[Ans. PV 1 -➔ 2 = 0.99]