'Mtive Techniques) OOSlPtJ

~
. semester (Q uanu ....
J',1BAFlfSt

90 (Unit I) [ chllpter s],

Regression
'[bi
•f1i . nee between
01 ere .
Correlation tW<
5·13
mlld#ddi1-11•1,r?$'\,.·a-1t1
· · · AnalysJS reg
and Regression ·tterence between Correlation llnd me
Table 5.1 shows the di
)'l'o
, I ... , ' • • ,

Regression: t een Correlation and R esslon

Meaning ,. ,_,, to' calculate a be w Re ession in<
5 •11 • • . . · · al tool usc:u . Table S.1: DUTerence Correlation . . al
varlo~s independent lin
continuous de~ndent van~le fro:
Regression analysis IS a s~IJSIJC
for ' pi~ction and
variables and IS commo . y u~\1turi r''!cnows that the
I Base linear It is a statistic too)
eaning Studr o~ th\etween used to calculate· . a
relation~hip .s 1cnown continuous · dependent
1
forecasting. For ex~mple, 1f agnc rel!t~d' then he will two van~les variable ft:om various
yield of rice and ramfall are clo~elr . ' ired ·to a1=hieve a as correlauon. inde endent variables. .
want to know the amount. of ;-un. requ )1 t • will use the are One variable is random
certain production. In this _s1tu~tio~-~ .modelling and ~hies Random variables variable and ~is
re ssion analysis. 1bis technique is u r dent used. fixoovariable. It may be
greal . f mathematical data using the value of a depeD
an J.SIS o , bl ,D pendent possfule that sometime
varuible and one or more independent varla ~ e sin le both the variables are
. bl also known as respome wriable, is the g
v~able, redicted by the regression model. Independ~nt random. . -..
vana e P
variable, also known as predictor variabll;, is
• used t predict
~ . It does not indicate the !t s~ows the ~ ~
cause · and effect eff~ ariabr:elationsbip F
the values of dependent variable. Regression analrs1s IS the
rocess to calculate the unknown value of one vanable from · · hip - between between
re1anons , . v . fu1es ; and a
p 'abl . variables and gives the creates a nctional t
the known values of the other van e.:, , ,~ ~
de ee of relationshi : . relationshi ·. ..
According to Bla_ir, ''Regression is · the measur: of ~e u Used to test the U~ _for prediction and
average relationship between two or more ~anable m ~ relationship , between forecasting, -,
terms of the original units of the data." \ two variables.
According to Taro Yamane, "One of the most frequently Valu Range It is a relative measure It is ~ a~~lute value.
used techniques in economics and business tesearch, to and its coefficient is lies It 'is- possible that the
find a relation between two or more variables that are between - l -to +1: coefficient ·· 'is greater
than 1.
related causally, is regression analysis." \ .
ematical Not much valuable for Widely used for further
Regression analysis makes an equation to desc~e the Treatment fufther mathematical mathematical treatmenL
statistical relationship between one or more pf~·ctor treatment.
variables and the response variabie ..It helps .to know, how Dependency Dependency of Its shows the functional
to determine. the nature and strength of a relaton 1hip of Variables variables on each other relationship between the
between independent and dependent variables. Regres~~n is not im ortant. two variables.
may be classified as follows: 1
1) Simple Regression
2) Multiple Regression
5.1.4. Regression Lines
3) Linear Regression The_ lines of the best fit between two variables and
4) Non-linear Regression
statmg .the c~mmon average relationship are known as
regressi~n Imes. There are two regression lines for
two vanables.
5.1.2. Application of Regression
Analysis When we take two variabl X
Following are the basic uses of regression analysis: Y as independe t . · bl es as dependent variable and
1) Used to kno~ the relationship between different (one \ X
n vana e then the tr . ht line 1s
as the regression li f s aig
. all d
c e .
nd Y and vi~e versa- is also'
or more) yanables. . 1 •,
true. The regressio:e Jne ;
0
2) Uspd t~ find out the coofficients of correlation (r) and possible values of y t th . Y on X gives the finest
cooffic1ents of determination. or e given values of X
I I •

3) In corporate sector it is useful to check the quality. I So this line is the sum of
1
4) Also very useful to determine the stati ti al 'c'.11~ulated values of y , - squares of deviations of the
(demand, supply; etc.). s c curve nununum'; i.e., and 'observed values of Y is
L(Yc - Yo)2 = minimum
1iques) GGSIPlJ
f{egres. sion (Chapter 5)
91
w11ere,
Y := Calculated value ofy
. . used to detenrun
· e the best-fit line
..,'_Observed
,o- value of y This technique 1s .d for
al
J>loW, for the regression line of X data by minimising the _sum .of sdqualaredse~~~:i:;e~e:~
1:(Xc -Xo)2 - nun be tween th e ac tual and predicte v ue ·
-
· ·
unurn.
on Y: is used to minimize the sum of squares, taking the .
residual
t
;sion .
w11ere, x, =Calculated Value of X
Xo = Observed value of x
. d . .
as the partial enva~ve_ 0
f the cost function w.r ..
artial derivative
Coefficient of detenrunation, set the P ffi . t OLS
equal to zero and solving for each coe icien · d f
....,is is why, to understand the co technique is also known as leaSl squares metbo or
relation '" . 1· ncept of
1wo regress~on m~s. are required. If re~ession line, regression or linear regression.
regression lme eJUsting between two ~e~ 1s only one
,Iation and ,neans r = ± I. anabJes then it . . . .
8 y nuIU11UZJ.ng th e sum of the squared residuals, the
al d b
parameters of the regression model can be ev uat~. _Y
.Ob ..row If the value of r is o, the bo Ordinary least squares, or linear least squ~es. A line is
'J i• d Jin . n th the .
sioo independent an es will cut each other at . vanance are drawn joining data points by using this technique _and these
sticaI tool Jines will be parallel (II) to x-axis and Y- . )a nght angle (i.e., data points minimizes the sum of the squared 91_fferences
Y axis (figure 5.1). between the obtained values and the correspondmg fitted
lculat.c. • a
dependent ri r=o values. OLS tecluµque is a commori technique for
I VariOUs computing the coefficients of linear regression eq~atjons. It
.riables. is used to describe the relationship between one or more
s random independent quantitative variables and a dependent variable
- . - _I
Q!ber..is i.e., simple or multiple linear regression.
rt may be 0
X
:ometime
Figure 5.1: Two Lines of Regression 5.1.6.1. Assumptions of Regression A~ly~i~
bles are Using regression analysis, one make certain assumptions

.use and s.1.s. Regression Equations to make valid statistical inferences. The different
applications of the simple linear regression models
tionship · Regression equation is u~ed to show the regression line depends upon the many assumptions which are as follows:
.es and and there ~ ~o regres~1on equations because there are ~.!_near _!egression _is . a rel~tionship _ ~t~,een .
1ctional two regresston hoes. So if_ there ~ two variables, say X independent variable(x) and dependent variable (y)
and Y, then the two re~s1on equations are written as: and its equation1s gi·ve~ by: -
on and I) R~on Equation of X on Y: The regression
equation of 'X on -Y' is written as follows : -.-#--
y=a+ bx+c
Where c = constant that represents to the deviation of
X=a:tbY - specific value of y and its desired value for a given
value.
at the To calculate the values of a and b, following normal value of x. It is used to describe the relationship
:reater equations are solved simultaneously: between the v_ariables x and y.
DC=Na+bIY
2)1\ Linear regression gi_ves_an e~pected valu~of v¢~le
urther DCY = aIY + bIY2 V y and these are normally distributed for each value of
nent. 2) Regression Equation of Y on X: The regression variable X. This expected value of normally
equation of 'Yon X' is written as follows : distributed values and it is situated on the regression
ionaJ
Y=a+bX line.
t1 the I
~ The value of x is fixed but the value of y is a
To get the value of a and b, following nonnal
continuous random.
equations are to be solved simultaneously:
1)1 It is supposed that sampling error associated with
wd
LY=Na+ bD(
LXY = aDC + bLX2
mean values of y. It is an independent random
variable and distributed normally with mean constant \
for
as In the nonnal equation DC, LXY, rx 2
and IY givthes
· d from ,e
2

?17
standard deviation and zero.
The standard deviation and the variance of the
the summations and can be detennme
· bles 'X and Y · expected values of y are constant for all values of x
observed pairs of values of- tw~ vana
within the sample data range.
N is the number of observed pairs of values.
nd ,J For the value of x which is outside the boundary of
ed the sample data, the value of y cannot be calculated.
so '
,i! S,1.6. Ordinary Least Squares (OLS)
:.&t
!I
Regr • 5.1.6.2. Computation of Regression
. ess1on linear regression Coefficient
1i
~ regression testing is known as th~ the unk"iimrn
rn~od which is used to de~rmt~~ totally depends
If 'b' is the slope of regression line of 'Y on X' , then bvx
IC
Pai-drneters "in a model. This technique is known as coefficient of regression of ' Yon X'.
are residuals
~ Upoij" Ifie mmuruzmg the sum of th e sb
dinary Least So, byx: =Regression Coefficient of Yon X = r ~
~!ween the actual and predicted val_ues .r r:gression. Ox
quare~ regression is often known as hnea
 ntit.0tivc Tcch nic1ucs1 (JOs; T
. seme ster (QUII
MBA pirsl ..
'"1
k for both log1!lt1c r e ~
. formula t,cst wor r:ssio ns. We ~an har
This Jtiple linear reg fficients with dly~
92 (Uail I) pr0p 0ni~
' Of ·x on y• . and mu
the standar I
dized coe ..."'-I
Similarly, for Lhc rcgrc ssion Jme CJx
X n y == r - mod es.
. coefficierl of hazar s d .
ffidents: Tots type of COeff~
o CJy
bxv = Regression Unstandardl1~ C(,ethe independent v~ab
2) . enerated using le after .
J{cgression of linear regression mo<1c1 Ill
of is .g . ·th the help . s which ~
one.-• l scale
trruning wi •aina
5.1.6,3· Prop ertie s these are rneasur I
ed ·n their · ~ Z)
Coeffi~ie~ ts. f correlation is. calculate d by t is talce in the
its in which we have to take,~:
same ~~rce
1) Coeffi~1e_nt oh geometric mean ' of the two I to train the model .;,_,
dctennmmg t c dataSet fro01 the ffiso..... i·ent does .n otelirnin'ate.the unit•'lt
. · dized coe
. n coefficients i.e.,
regress10 unstandar . hc uld not be us ed to drop or ra.1.Of
measureme? so Jt dent s O
t
variables). -..."
r=± /bxv xbvx . d 'ts value is redictors (mdepen
P . .
WI re r - Coeftkient of co rrelat1 0n an 1
. coefficients are want to estim ate the tncorn e ('
;hen both the regression I so the value For exam Pie• if webase d on its age ("m
i~ .
pos1t1ve r years), heig1ii1n
positjve. Because r ~es betwe en +I to - ' rupees) of a perso~ ht (in kg), then
the inputs to
. t more than one. (in cm), and w eig age heig ht and weig
of regression coefficient ~s _no ut not from scale. . n mode1 are ht, and •~-
regress~o. ' .
2) It is independent from ongm b e We have the follo wmg ~ic
. . nt is greater than unity • then
coe ffi1c1e output is incom( ·pees) = ao + a1*A
formula:
3) If one. regre
.
ssion . ge (years) +
remammg must be smailer than unity . . . less than the
Income ru
. ht rem) + a3*W .
e1gh (k ) (1
The value of coefficient of co~J auon a2* He1g , t g + e ... )
is
4) mean value ofregression coefficient ·• ..·, this for expl ainin g the effec
s. t of each
yi,e us~ t varia ble on the outcome
s.t.6.4. Standarised and Unstandarised mdepen ~,n
(respo nse outpu ·
t) Its inter preta tion · is simple and
. d All othe r vana . 3)
Regression Coefticie~ts C ffi . nts· Stand str.aighftfolr bles are constant and a
• darised Regression oe 1c1e • arised w~t: X·(p redic tors)
1) Stan . shift o uru m , mea ns that there is an
oeffi . ts are the regression coefficients )
regression c 1c1en
which is multiplied by the standard deviauon . ~ th t average c h ange o f a·' units in Y (outc ome s .
or d ad In this example, if we supp ose all data are
independent variable and divided ~y the statistically
.s:n ~s . . 1cant like a - o 3 a 2 = 0.2 and
deviation of the dependen.t vai:iab~e s1gm 6 a3 = 0.4, then
an
(Standarised regression coeffic1en_ts) is given .
1- • , .
by. these coefficients are inter prete d as:
Standardised Regression Coefficient
Standard Deviation of Wh en · w e · assu me that othe r . varia bles are constant
then getting 1 year olde r is relat ed wt.th .
. an mcrea~ of
Independent Variable o 3 in income. Othe r varia bles are heig ht and weight
= Regression Coefficient x Standard Deviation ;hic h are cons tant. If we chan ge "the
of independent 4)
Dependent Variable variable by one unit keep ing the othe
r independent
variable cons tant, it deno tes the amo unt
by which the
The coefficients cannot be compared depe nden t varia ble chan ges.
with ~ e
magnitude of the coefficients of the indep
endent
variables, which are measured on different 5.1.6.5. Decision
scales till Ma kin g Based
one standardizes coefficients. If anyone
divide
on
independent variable by number 10, it will Regression Analysis
increase The helpful statistical meth od that can be
coefficient by ten. lever aged across an
-· organization 'to find the degr ee at
This division does not change the association whic h particular
between independent variables are affec ting depe
that variable and the outcome. However, if nden t variables are
the variable by number 10 then this will
we divide known as regression analysis. The diffe
rent scenarios which
5
not only are. used for c~nd~cting regre ssion anal ysis
increase the coefficient by tenfold, but
also will prov ide valuable. E
increase the standard deviation. When we multi actionable busmess insights and these are
ply' tbe endl ess.
coefficient by the ratio of the standard devia For runni~~ smo oth and effic ient orga
tion of the nisa tions we require
independent variable to the standard devia bette r dec1s1ons and it mus tund ersta nd
tion of, the d .. the ·nflu ence of the
dependent variable, we get the coefficien~
having bot ecist on taken . Data of sales , inve stme 1
units. As the different independent varia nts, expenditures
bles are on and othe r para mete rs are colle cted by
the same scale , one can compare these varia the Orga nisations. It
bles. can als? be anal yzed for impr ovem ent.
With the help of
If all independent V3?ables are present on regres~to~ analy sis, orga nisa tions mak
the same then 1t 1s used to e sens e of the data
scale and also the vanables are dichotomo R · . . . .
us then th . gain ms1g ht into an orgamsanon
standar~zation will .be unnecessary. The e~res~10n anal ysis is used by busi ness .
do~n side ~ anal ysts and data
s~da rd1z ed coefficient can be defined as 0 pro1ess1onals to mak e str t .
it is the one 1) Optimisation of .
which loses a sense of the actual effec a_egic busm ess deci sion s: ..
"ble h t that h anal · •
vana as on the dependent variable. eac Busmess: The purp ose of regre ssion
out c~:; s ~ ~sf o?D .colle cted data into action
able
· garu zatto ns are embracing data-driVCll
 . _.<,n (Chapter 5)
~rFn·~~1 93
¾>u
ciecision making_. This helps in rerno .
ell as guessing or assum·
mg a h
Vtng old techni
ques Ordinary Least Squares(OLS) Fit
sll dri
uJtirnateIy . ves perfonnance . YP<>thesis that 5.1.7.1.
practical assistance to the rn lJl an organizati of Linear Regression • f x 00 y To detennine
.n;7ation is provided by thianagernent unit of othn. + by be the 1me o ·
org"'- s anaiys· . e l) Let X = a d b the following two normal
11vailable data, ~e data can be an is. With several the values of a an , I
e11~ . e to be solved simultaneous y:
understood to gam efficient insights alysed as Well as equations ar
. ill! . . and Work smaruy . LX =Na+ bLY 2
cllt<i predictive Analysis: For pred· . DIT = aLY + bLY
ans. 2) rganizations use regression an~c~ng future event be the line of y on X. To detennine
tlle o e business analysts predict thYSis. In this proces~
th 2) Let y = a + bX d b the following two normal
I'he vana . bles fior given de outcomes of the the values of a an ' . ultaneously:
dependent a · ependent v~a · b.Jes. equations are to be solved st m
t Of for diuerent unportant purposes like f,
volumes or create growth plans et orecas~g sale LY=Na+bLX 2
llllc
c., the multivariate DIT = ai:x + bl:x
ll·near regression is· used. Foll~w·mg are the general
fi
procedure or usmg regression t0 . data obtain the two
(in make good Example 1: Consider the fo 11owmg
predictions:
~ht regression e~q~u'..'.'.a~ti~on~s::_:.--::---r-;-~,izi 'Cl
i) Do research on subject-area so one c
I~ I~ i ~ ~
to 3 1
the work of-others. It research :111 construct 4
:he sub steps. con s1sts of other 12

ii) Gath~r data for _the important -~ilriables. Solution:

iii) Specify and estunate the regression d.·
dlth fi moe1, ., Calculation of Re2ression E;quafions
iv) I~ a mtho e ~t its well enough to the data is X y X' yz XY
:::h given en use 1t to make predictions. · · 25 64 40
5 8
:ie
3 12 9 144 36
ld J) Risk Analysis: At the time of data analys~s.' mistakes 10 6 100 36 60
a are made by the analyst and confusion arises between 8 5 64 25 40
lO correlation and causality. It is significant to note and 9 14 81 196 126
understand that correlation is not causation. · 2
l:X-35 l:Y-45 l:X: -279 :EY -465 l:XY=302
2

Organisations use regression analysis to evaluate the

:y risk in the financial domain and other such domains. Wh~re·N=5
:n Thus it guides to make important business decisions. Regression equation of X on Y: X =a+ bY and the two
The capital asset pricing model is developed by the . normal equations are:
lt linear regression analysis and it helps in taking LX• =Na+bLY
,f financial decisions . ., : LXY = iu:;y + bLY 2
tt
4) Understand 1 the Failures: In addition to data analysis, Substituting the values
'. t ..... (1)
the regression analysis is helpful in an organisation to 35 = 5a + 45b
understand the failure and learn to correct it in future 302 = 45a + 465b ..... (2)
analysis. Generally, this regression analysis offers
Multiplying equation (1) by 9:
quantitative support for the decision-m~g pr~es~.
The one of the main properties of regression analysis JS
315 =45a + 405b . .... (3)

predicting success. For example, analyzing current and Subtract equation (2) from (3),
previous sales data in an organisation, one can 13 = - 60b or b = - 0.22
understand and predict future success.
Substituting the value of b in equation ( 1);
35 = 5a + 45(- 0.22) ⇒ 35 = Sa - 9.9
5.1.7. Determination of Linear ⇒ Sa= 35 + 9.9 = 44.9 ⇒ :. a= 8.98
Regression Equation Putting the value of a and b in the equation,
To determine the parameters of a fitted regression The, regression equation of X on Y:
equation following methods are used. X = 8.98 - 0.22 Y
Substituting the values from the table in the previous
lDetermination of Linear Regression Equatio~
exa'inple,
- Method of Least .._ . Regression Equations when
- 45 = Sa+ 35b
302 = 35a + 279b
..... (1)
·· ··· (2)
- Squares
-
Deviation taken from
Act:ual Mean
..... . . . .
..
Multiplying equation (1) by 7,
315 =35a + 245b ····· (3)
.. .,, ' ...
R~we~~ion Equatj ~n.;5 ~J1.en. , Subtracting eql)ation (2) from (3),
Dev1aiio11· taken from tbe 13 = - 34b orb= -0.38
'Assumed Mean -
 (Quantitative Tccbniqucs)
~
1: .
MBA FiJSl SemeSld" . data obtain the Re

94 (Vair n . tion (I), . m the followmg

T:
. value of b ID equa Eum 51 73 Ill S7
Substirunng tbe (- 0.38) ti6 re ~:::_.,;;,; •~ 39 61 80 47 6(
=
45 5a + 35 - 58.3 or a 11. = 7
6'.
5a = 45 + J3.3 - .
• d b in the equanoo . equatiaos refei- table 5.l.
··gregres.goo 3'
- lbe values ofa an
Putting Solution: To obtainiD
b
• ofxandy:
Regression equabon
. n of Yon X:
Regression equaoo O 38 X s
y = J 1.66 - - the and Y· 71
Letx=a+bybethe line on X
.
. n equations from 7
Example 2: Find the twOthe re~~~ of X, if y is 16: To find a and b• we have •'--
~ 1~ouowing
'
normal equabOQs: a
j
following data an d ti.mate
4 / 5 / 6 / 12 7 l:x=Na+bl:y
/ X / 3
I v I 7 I 6
9 II l:xy = al:y + bl:y2

Substiruting the value. we get

Sol D tion·
Table: - 900 = 10a + 700b ··--0)
X y x· y• XY - 66,900 = 700a + 51,868b ··•-(2)
3 7 9 49 21
Table 5.2: Calculation ol R
4
5
6
9
6

11
16
25
36
36
81
121
24
45
66
s
91 - 71
.
-- 6.461 8.281 5,041
7 49 12 144 84 71 .J -, 75 7,Xl5 9,409 5,625
LX-25 L\'=45 Ul=135 I:¥1 -43} 1:XY=240 -7,452 11,664 C
108- -;- - 69 _. 4,761
121 97 11,737 14,641 9,4()1} e
WhereN=5 67 70 4,690 4,489 4,900 0
124 - ~91 - . 11,284
Regression equation of X on Y,
51 39
15.376 8.281 s
Let=X=a + bY ..... (A) 1.989 2,601 l.S21 p
73 61 4,453 5.329 3,n1
I' Then two normal equations one 111 80 8.880 12,321 6,400 l

~ II
LX=Na+bl:Y ..... (1) 57 47 2,679 3,249 2,2()IJ l::
LXY = 31:Y + bll2 ..... (2) I:ll:=900 l:.v=700 =87 =~ a
s
Substitute the values from table in equation ( 1) and (2) WbereN=IO
25 = Sa + 45 b · · ... (3 )
I 240 = 45a + 431 b ..... (4 ) Multiplying equation (1) by 70, we have
Solve equation (3) and (4) 63.000=700a+49,000b -' ·---(3)
a =-0.1923 b = 0:5769 Taking equation (2) and (3),
Putting the value of a and b in equation (A) 66,900 = 700a + 51 ,868b
X =-0.1923 + .5769 X16 63,000 = 700a + 49,000b
X =-,0.1923 + 9.2304 3,900 = 2,868b
When Y = 16 then X =9.0381
b= 1.36
Now Regression equation Yon X, Substituting b = 1.36 in ... equation (1),
Let Y=a+bX
... .(B) 900 = 10a + 700(1.36)
Hence the normal equations are, 900 = 10a + 952
1:Y=Na+ b}:X a=-5.2
LXY = a De+ bD(2 ..... (5)
..... (6) :. Required equation of x and y:
X=- 5.2 + l.36y
Putting the values from table in equations (5) and (6)
45 = 5a + 25 b ..... (7) Regression equation of x and y:
240 = 25a + 135 b ..... (8)
Solving equations (7) and (8) Let x = a + bx be the line on y and x
a= 1.5, b,=: 1.5
To find a and
Iy:::Na+ bixb, we have the following nonna1 equations:
Put these values in equation (B)
. Y = 1.5 + 1.5 X 1 Ixy ::: al:x + bix?
Hence, the two regression equations are, Substituting the value, we get
700 ::: 10a + 900b
X =- 0.1923 + 0.5769¥ and Y = l.5 + 1.5 X
.....(4)
66,900 = 900a + 87,360b
And when Y = 16, then from 1st equation, X = 9.6
..... (5)
Multiplying equation (4) by 90, we have
63,0()()::: 900a + 81,000b I
.....(6)
l
l ~c.!,!rcs~ion (Chapter 5)
95
rating equation (5) and (6),
66,9()() = 900a + 87 ,36Ob When, X = 7 ~elirs then, y = 69.7 + 1.13 X 7 = 69.7
63,000 = 900a + 81,0()()b + 7.91 = 77.61 ..
3900 = 6,36Ob
Example 5: Derive regression lines for the follov:-ing data:
b:: 0.6 1 LX = 30, txz
~d90, LXY = 192, LY = 30, LY2 = 190,
substituting b = 0.61 in equati N=5.
700 == IOa + 900(0.61) on (4 ), We get
7 00 == IOa + 549 Solution:
a== 15.l l) Regression equa~on of X on Y: X = a + bY and the
two normal equations are:
:. Required equation of y and x· l:X=Na+bl:Y
y:: 15.l + 0.61x . 2
l:XY = 81:Y + bl:Y
Example 4: The following d t .
•
rnacrunes a a gives th
operator and their pefti Substituting the values
e· experience of ...... (1)
b number of ood arts out of lOoO~ance r.ating as given 30 = 5a + 30b
1eces· \ 192 = 30a + 190b ..... (2)
. I 2 3 4. 5·
o erator e enence -l ' 6 7 g
Performance e erience
16
87 -8; · 9 68
18 4 3 IO 5 12 Multipl~ing ~quation (I) by 6:
180 = 3pa + 180b ..... (3)
. -8 1-8 80 75 83
Calculate the regresSJ.on line of rti . I .
Subtract eq~~tion (3) from (2),
experience and estimate the Pr:ab~~a:e ra~ngs on
- -12.~ -1.0!? or b = 1.2
operator has 7 years experience. -~._, pe1or¥fi~ce of
Solution: Let Operator experience b ct f ; Putting the value of bin equation (1); , ,, .
perfonnance experience by Y. e e-noted b)! X and 30=5a+ 30x 1.2⇒ 30=5a+36 ⇒ -~~ ~-·:;-,6
Let Y = a + bX be the line of y missin al :. a =-1.2
.
binomial expansion on X. To deternun~ve thue byalusmg the
th fi II · · e v ues of a Putting the value of a and b in the equation. The
and b, e _o owmg two normal equations are to be s I d regression equation of X on Y:
simultaneously - 0 ve
X = - 1.2 + 1.2Y.
1:Y=Na+bD{
2) Regression equation of Y on X:
LXY = a~ :+- b1;X2

I
To determine the values of a and b the following two
Cal C iii°8 tiODOfRe21"ession Equations ·.,normal equations are to be solved.
X y xz XY l:Y =Na+ bl:X
16 87 256 1392 l:XY=~+bD(2
12,,-. 88 144 1056 Substituting the values,
18. 89 324 1602 30 = Sa+ 30b ..... (1)
4 -' - 68 16 272 1'92 =30a + 190b ..... (2)
3 78 9 234
10 80 100 800
Multiplying equation (1) by 6:
5 75 25 375 180 = 30a + 180b ..... (3)
12 83 144 996 Subtract e.quation (3) from (2), 12 = 10b or b = 1.2
.tX-80 l:¥=648 EX:2 =1018 nY=6727 Putting the value of bin equation (1);
Where N=8 30 = Sa+ 30 x 1.2⇒ 30 = Sa + 36 ⇒ Sa = - 6
The values, :. a =-1.2 '
648 = 8a + 80b ..... (1) . Putting the values of a and b in the equation .
6727 = 80a + 1018b ..... (2) Regre.ssion equation of Y on X:
Y = - .l.2 + l.2X
Multiplying equation (1) by 10,
..... (3)
6480 = 80a + 8Q0b Example 6: Find equation of regression line Y on X for
..... (4) .
6727 = 80a + 1018b the followin G\ata:
X 25 23 28 26 29 21 30 24 33 36
Subtracting equation (3) from (4),
Y 28 26 28 24 29 26 31 32 36 27
247 = 218b orb= 1.13 Also estimate Y, .if X = 55.
Substituting the value of b in equation (1), Solution: Regressf on equation of Yon Xis: Y = a+ bX
648 = 8a + 80 (1.13)
8a = 648- 90.4 = 557.6 or a= 69·7 Tio detennine the value of a and b, the following two
normal equations are to be solved:
Putting the values of a an~.I;, in tbe equation IY=Na+bIX . . .... (1)
Regression equation of Y on X: IXY = aD( + b!:X2 ..... (2)
Y=69.7 + 1.13 X
 . Techniques) OOS(pt,
uantitattve
rnester (Q
MBA First Se . nt of Yon X = ¾.
. Coeffic1e
· 96(Unir I) ess1on X}:
Oy = The RegT ~X Y)
CalcuJatJon of R r-
~x -J ::E;xy ""'XY
o,- ,_,
~
~ s
X
_J;(x-XX_'.'-Y • ::E;x' - -
LX' Nxr
25
23 bvx- 1:(x-xf
28
26 24
·-X); y= (Y-Y)
29 29 Where, x. = (X
21 26
30 31 the following data obtain two
24 32 ExamP Ie 7·• Usi g
33 36 20 27 21
36 27 41 39 20
LX=275 .D'=287
- R ess1on E.!luations
I -X- ~ Y= ~
WhereN= 10 . . Cakulation ' Y-
Solution. y ,r
:ty
Substituting the values, .. ... (.3) I V y , X- -

287 = 10a + 275b ..... (4) , X / X Y +--;:;3~6- , :44<;91 I 4422 7

7951 = 275a +
7757
b . tion (4), -l-----:321
1 -T -=J6~;;- :::7t...·-f--=-;t-r.~4!J ~4f -7
. (3 ) an d10mequa I
I ·. 1
1;!_5'-f--;~ ;-;T =2~
36 - -;-; _. 2 4 100 30
10 9
Multiply 275 in equatJOn . . . ... (5) 109 I ' , 1

~ ~
78925 = 2750a + 75625b ..... (6) 48 ·3 . . 0 I 0
79510 = 2750a + 77570b 1
7
. . (5) from equation (6),
Subtractequaaon
I
25c:
; : · 3.
50 ': , · · ·1~
7 +----'1-;--I6-:--7~,1~c,· I 4118· I
22 45· ·5-
585 =1945b ⇒ b =0.3 -1. 25 25 25
1 16 33 · ;, ' 5 - 1 9 -3
~ 3I
⇒ ·41 •,; .~'f.
. fb · equation (3), 1
20 36 1 6
Puttingitlie value o m 287 = 10a + 82.5 6 1
287 = J'Oa + 275 X 0.3 ⇒
27
'· 21 39 18 O 324 0 r
1Oa = 204.5 ⇒ a = 20.45 I

LX= 20 0 LY O LX2 LY-= LXJ=

~v ,~-o
. equation of yon X is:
So .regress10n 706 , 159
I ., 210 ~80 - 128 i -
I
¥=20.45 +0.3X
I

'
Wh X = 55 then, WhereN=::-10

L
en, y = zo.45 + 0.3 x 55 = 36.95
- 2~0-- 21 and y == 380
10 =38
·X- -fo-.-
5.1. 7.2. Regression Equations . w/b.en

!!e 11
. ti Taken from Actual Mean
onfiormulae are used to find out the deviation of X
LXY ~ 706
= b y2 1 -
159 = 0 225

t
vi~ bXY :
.--o owmg
and y series when deviation is taken
fr th tual mean
om e ac 7 1 .
l) Regression Equation of X on Y / I '"xy 159
b
YX -,Ix2
I -. . :;::}.242
- ~ =128-
X-X=rGx(Y-Y)
Gy

X = Mean of X Series and Y = Mean o · Y Series Reg,elsion ~uatiQll 0f X on Y is given by

X- XI:: hxY(Y - Y)

⇒ Ix -21
r CJ'x
CJ'y
= Regression Coefficient of X OJ! Y or hxY
1
=s 0.225(Y - 38)
I(x-xxY-Y) Ixy 1 ~ X ~0.225Y - 38 x 0.225 + 21
bxv- 1:(Y-Yf =LY r
O
1.~ ~ = 0.225Y - 8.55 + 21
) ==-
LXY- (rxXLy)
_ry2_ ~:,
Where, x =CX-X); y = (Y-Y)
I ⇒ X = 12.45 + 0.225Y

Example 8: An Investigation into the demand for LCD's

i,n seven towns resulted in the followin data:
Po ulation <Lac) LCDs Demand
2) Regression Equaljon of Yon X 11 15
I 14 27
O'y f. -)

I
-
Y-Y=r-\X-X
O'x 14 27
17 . . 30
X == Mean of X Seri~ 17

and Y == Mean of Y Series 38

.25 46 [

:iruj,le~on equatio? of Y and X and e s ~ demand the r

W Leos for a town With a ~ of 301housand.
, -- I
 97

, sion (Cha pter 5)

f(C~ rc-

5ol1Jtion:
Popu latio n X xy
""' CX-t 7) y
Y y=(Y -31)
II
14 (LCO s Dem and)
- 16 +96
15
14 -4 +12
27
17 -4 +12
27
17 0
30 -1
21 +3 0
34 28
25 38 +7
......, 120
~ 46 +15
"""" ' ll9 L.X: :O
~2~68
--= --~ ~- ~l:Y!:..==l2:!'.17~__LJI:~=~OLJ~~~~l'.:=
Where N =7 ·
11 9
xo::Y-
N
=-7 - = 17a ndY =~
N =7
217
==31

r,iow prepare table . The n, b • ::: L xy _ 268

y Lx 2 -13 4::: 2

Therefor:: the Reg re~i on equa tion of y on X :

y _ y = by, (X- X)

substituting the know n valu es

Y - 31 = 2 (X - 17) ⇒ y ~ 2x _ 34
+3I ⇒ Y==2X-3

To obtai n the aver age dem and of LC Ds whe n the popu Iat.10n . . 30 thousand , we substitute 30 in place of X in the
1s
°
. equati on f y on X , then we hav .
regression
Y = (2 x 30) - 3 = 57 LCD s
e,
..
Example 9: From the follo wing data find
~o¥~r
X~tw
~~ ¼
~~re 22s~:
16 et 18 26n~li~ne~
s~s1~o~ i 434~lJ~29~=3311o=c~D
20
41
y 32 22 48 38 52 46

Solution: Calculation of Re2ression Equa tions

y Y) x2 y2 xy
X X (X X) X (Y
--6 - 7 36 49 42
16 32
17 16 289 68
18 22 -4
16 81 36
26 48 4 9
0 1 0
22 38 0 1
144 169 156
34 52 12 13
7 49 49 49
29 46 7
121 36 66
33 11 --6
II 4 4 -4
41 -2 2
20 :Ex2 -386 :Ev2- 678 Lxv -413
I:x- 0 1:v=O
LX- 176 1:Y -31 2
Where N=8
312
X= 176
8
=22 and Y=
8
=39

Ixy 413 2:XY 413

bXY = I l = 678 =0.61 and byx = ~)2 = 386 =1.1

n by
Reg r~io n equ atio n of X on Y is give
X-X =b. xy ( Y-Y ) .8 + 22 ⇒ X =-1 .8 + 0.61Y
Y -39 x 0.61 + 22 ⇒ X = 0.61Y -23
⇒ X-2 2 = 0.61 (Y -39 ) ⇒ X = 0.6.I
n by
Reg r~io n equ atio n of Yo n Xis give
Y-Y =IJ dX -X) - 24.2 + 39 ⇒ Y = 14.8 + 1.lX
:::> Y -39 =1.1( .X- 22) ⇒ y = 1.1 X- 22 x 1.1 + 39 ⇒ Y = l.IX
find correlation ,coefficient from the
10: Obt ain the line s of regr essi on and using regression coefficients
~p ie
9 15 19 26 31 35
5
ollowing data: 39 48 59 65
No. of Service (in ears) 22 26 35
MonthJ Income (fin Thousand)
 MBA f'ust sernesrer
(QUantitative Te ch
ni~ ) %ti ~
9s (U nit n 1111
fle gJ'
Ca)c:oJadOO of _ xy
Solution: _
SerVic:e Monthly (nc:Om
y~ e ... ex- X) S='
_ --1--
-20~~-t -,
C -~s_--J------;~---r= 176
35
9 26 i6
-:.!..7-+ -- 3
L _j 1! 25 --- +- ~3
19 ~5
39
-- r= ·==3:...---r-- 36
26 .6 ~- +- ;-; ;.- 187
48
31 59 ~7:...--1----;; 345
3 :1 08 2
35 65
~= 14 0 IT = 294 =
W he reN =7 Now find out the value of r~ by ap pl yi ng the followin
<Jx
&
X= 1~ =20 2
and Y= ~ =42 formula :
~
"""
b,cy =£.. 1082
--,X'=f -= - ~xy = 1082 l.43
Ly2 15608 .6 9 , bv x- "" " 2 -75=
4
£.,,, x
E"-2
an d
Regression equatio
n of X on Y is given
eq u:
by: ."' :A d .= y ._ Ay
d, = h -..;
~
X - X = b,cy (Y W he re,
- Y) __ - ~ - Y --
⇒ X -2 0= 0. 69 (Y -4 2)
⇒ X = 0.69Y - I
42 x 0.69 + 20 Exam~ e ;,. 11 . -Fr om th e fo llo wi ng da ta fin
⇒ X = 0.69Y - regression ...,,.,~ ation d the
28.98 + 20 ,, . . ·s-• an d ~s tim at e th e lik el y va lue of '( Soll
⇒ X = - 8.98
+ 0.69Y · when X = 10 0. rn ai
X 74 76
98 82 58 76 94 88 5 as st
Re gre ssi on equatio y 124 131
n of Y on X is giv en 117 131 96 l
by: 121 136 CJ7
y - y = bYX (X - X) 85
⇒ Y -4 2 = l.4 3( X- 20 Solution: Le t assume

~
) th at th e m ea n fo r X-
⇒ Y = 1.43X - 20 x 1.4 for Y-series is 120. se ries is 80 and
3 + 42
⇒ Y = 1. 43 X- 28.6 +4 X
2 Y d:s=(X d, =( Y
⇒ Y = 13.4 + l.43X -A J -A ,.) dz
1
~14 - - 124 -6 5
Correlation coefficien 4 -2 4 36
t (r) ,9s -: - 1 31 18 11
16 7
=J bX Y xb YX =J 198 324 ll
0.6 9x l.4 3= J0 .98 67 =0.993 11 6 . • 117 121
-4 -3 12 l '.
i_ gz · . --
131
16 9
5.1.7.3. Regressio 2 11 LX
n Equations 22 4 121 4{
Deviation Taken fr when 58 96 -2 2
om the Assumed Mean -2 4 528 484
Ifwe take the deviatio 76 121 576
n from the assumed -4 Wh
following formulae me -4 16
are used to the regressio an then the 94 136 14 16
1
n equation. 224 196
88 97 256
Note: When Assu 8 -2 3
med mean is used
because value of 51 -1 84 . 64 529
ac tua l mean is in the 85 -2 9
fraction fonn. -3 5
LX = 1015 841 1225
Regression Equation LY Lds
of X on 697 =1038 td y :I:dxd, = td /= DI z_
Y is: =- 23 =- 42 1787 1 -
X -X = r 0 x (Y-Y) 1981 2854
a ,Y
W he re N =9

Now find out the va X = 697 -

lue of r 0 x by apply . 9 - 77.44 and 1038
Oy
ing the following y = -9 -= 11 5. 33
formula :
bXY = N Ld xd y -
L d xL dY _ 9x17
Ld d 1: d, X.Ldy 87 -( -23)(-4 2)
bXY or r crx = • Y
N L d; -( L ,d y) 2
cr (
N N Ld ,d y- "t" d "t" - 9x 28 54 - (- 42) 2
D m ~ •~ y
d = 16083-966 15
y l:d/-~ 117
Regression Equations
N
N Ld ~ -(L dY )2 25686--1764 -23922.
= 0.632
Re i
X-
of y on Xis:
Y- Y = r o-y (x-x) hvx =N ~ t: ;- Ld,~d, =9x 17 87 -( -2 3)(-421 IA '
Ox
x -( L ,d x)
- 16083-966 _ 1511 9x l9 8l -(-2 3) 2
7
17829-529-17300 =0
.8 74
 ....,;,,n l C hapter" 5)
f:<'/!n-
99
~ ion equation ofX on y • .
Jl · X-X=b,n,{Y-Y) JSgavenby,
· · ⇒ Y=
When Age (X) = 12 then maintenance cost (Y) 1s.
::::> ~ - 77.44 0.632(Y - 115.33)
= - 0.28 + 0.71 X 12 = - 0.28 + 8.52
::::> X - 77.44=0.632y _ 1 Y=S.24
15 3
::::> X - 77.44 =0.632y - 72.8: X 0.632
Example 13: Find equation of regression line Y on X for
=
::::> X 0.632Y - 72.88 + n .4,4 the followin data:
::::> X = 4.56 + 0.632Y X 28 43 48 34 65 52 70 61 25 24
Y ~ " G ~ ~ 00 ~ W W ~
\\fheo X == 100, then 100 =4.56 + .632Y
OO _ 4.56 ==0.632Y Also estimate Y, ifX = 80.
I~.4" ==0.632Y
~~ ~ 95_-44/0.632 =151.01 Solution: Let the assumed mean for X is 50 and the
assumed mean for Y is 60.
"'-·aJllPle 12: The following table . Cak:uladonofReeressionEouations
,:,., . gives the a f X y ct.= d,= (dJ' (d,)' d,.d,
and annu::tl mamtenance costs. Obtain ge o b ~ <X-A,) (Y-A.)
equations. AJso find out the mainte the regression 28 68 -22 8 484 64 -176
irho,e a!le is 12 ears old. nance cost of a bike 43 53 -7 -7 49 49 49
ve of Bike (in Years 48 42 2 -18 4 324 36
~ n112I-t-53~-::57:-4-,!,,;ll!_!.._!15U
1

~fs in tenancc Costs m 34 ~ -16 -35 ~6 1225 500

8 10 65 38 15 -22 225 484 - -330
52 60 2 0 4 0 0
Solution: Let X denote the age of bikes an
maintenance cost Let the assumed mean "' X~ y denote
70 62 w 2 400 4 40
. 5• ,or is 7 and the 61 20 II -40 121 1600 -440
assumea, mean tior y IS 25 80 -~ 20 6~ 400 -500
24 82 -26 22 676 484 -572
Calculation 0 f Rle """d on l .ouatioos :EX D' DI, I:d, I:(dJ2 I:(d,)1 I:d,.dy
Age of Maintmaoce d,. d =
1
=
(d,J~ (dy)'
d,.d1
=450 -530 --50 =-70 =2844 =4634 =-1333
Bike Costs (Y) (X - AJ (Y -A )
1
{X) WhereN=lO
I -5
2
5 j 3 -2
-4
-2
25 16 20 X= 450 = 45 and Y= 530 =53
10
.
4 4 4 10
7 I 5 0 0 0 0 0 - _ -50x-70
I:dxdy- I:dxr.dy
II I 8 4 3 16 9 12 - N
1333
15 I 10 8 5 64 25 40
bXY 10
2
= - 0.41
70
(5634)- (- >
I'Alz =5
~= / D ' =27 .Idy=2 l:(dJ2 l:(d1f l"Al,.dy=
=109 =54 . 76 .
Idxdy- IdxLly
10
_ 13 33- -50X-70
Where N= 5 ⇒ bYX N 10 =-0.65
Idx2- (Idxf (2844)- (-50'/
- 40 - 27 N 10
X= -=8· Y=-=5.4
5 5 Regression equation of X on Y:
Idxdy .r.dtl'.dy _ 5x2 X - X = bxy (Y - Y)
76
N 5 = 1.4 ⇒
⇒ X-45 =-0.41(Y -53)
⇒ X=-0.41 Y + 53 (0.41) + 45
I.dy2 - (l:.dy)2 (54)- (2)2 ⇒ X = 66.73 - 0.41Y
N 5
5x2 Regression equation of Y on X:
Idxdy .llixl'.dy 76-5 Y- Y = byx( X - X)
bn: = N =0.71 ⇒ Y - 53 = - 0.65 (X - 45)
. uix2 - (Llx)2
N
(10~- (5t ⇒ Y = - 0.65X + 0.65(45) + 53
⇒ Y = - 82.25 - 0.65X

When,
Regression Equation of X on Y: X g- X = 80then,
- -
X- X = bx (Y - y ) ⇒ X - 8 = 1.
4(Y 5 4) ⇒
- .
- -
4Y Y ::::: 82.25 - 0.65 X 80 = 82.25 - 52 ⇒ Y = 30.25
l.4¥-7.56 ~ X = IAY-7.56+ 8 ⇒ X=0. + 1.
44
Example 14: Find the two regression equations from the
. :
foll owmg
Regr~sion Equation of Y on X: _ S) ⇒ y - Year 2004 2005 2006 2003 2007 2008 2009 2010 2011
y - y = byx (X - ·x ) ⇒ y - 5.4 = 0.71(X Supply 70 72 76 81 73 75 79 86 83
S,4::: 0.7) X - 8 x 0.71 _ -0.28 + 0.71X Price 145 140 130 124 133 127 120 110 116
::;, Y= 0.71X-5.68 + 5.4 ~ y -
 MBA First e
S rnester (Qu
antitative TechniqUea) Cle.! ......_
ls1l\J ~
~egces•
·on (J), O
From equatlX+ b2 y + c2 ===
100 (Unit I)
. 80 and
the mean fo r X-series is a2 Regres
Solution: L~t as sume th at "-Y - C2 -b2
for Y-series 1s l;O. d, = x===-vz --
or a2 - b zy--
Cz ::::> b,y - az findo,
X (X-A.) x--
2
a2 i) M
70 145 -_ ~o or - a • . between X and Y can no...., U) St
•on coe ffic1ent r ~
72 140

76
81
73
130
124
133
- 4
I
-7
-5
The correlatlth.is formula:
obtained by
r2 ==byx. bxy- b1
)
-(3
(- b
X _ ·_2
a2
) ab
bl a 2
I 2
=
iii) C

sotuti
75 127 ") T
I 4

..... (4)
s
th two lines
of regression are:
WhereN=9 le 15:
ExantP J0=0 If e an d 20X - 9Y - 107· = 0 •

X= 695 =77.22 9 = 127.22

and -Y= 1145 4X - 5Y + se is the Jme . of regression of X on Y, and y
th
Which of e d CJ when CJx = 3.
X? Find rxy an Y •
9 " "d 9x(-644)-(-25)(65)
NLd.dy- L.A"-' Y - 9xl475-(65) 2 on .
01ven th e regression lines as:
bXY Nid:-(LdY)2 S I tion: We are o· - 0 ..... (1)
6 ou 4X - 5Y + 30
20X-9Y-1 -
0 07 - . . ... (2)
.
=-0.4 "'d "'d 9x(-644)-(-25
)(65)
NLd,dy - "-' •"-' Y - 9x301-(-25)2 and . f the line of regression of X on y
bvx NLd!-(Ld.)2 Let·( 1) be the equat~o; ~f the line ofregression of Y on X.
d (2) be the equatio 5
=-2.00 an 5 30 b =-
From (1), we get X=--Y- 4 4 ⇒ XY 4
Regression equ~on of X on y is given by,
X-X=b,o,(Y-Y) . _2ox 107 ⇒ bYX = 20
From (2), we get y 9 9 9
⇒x _ 77. 22 = -0.46(Y - 121.22)
⇒X _ 77_22 = -0.46Y + 58.5212 2 20
b . b = x~=2.7778
r=YXXY94
⇒x = -0.46y +58.5212 + 11.22
⇒ X = 135.7412-0.46¥
Since r2 > l ; our supposition is wrong.
Regression equ~on of yon Xis given by, [ •:we always have O ::S r2 ::S l]
ii)
Y-Y=~-X) Hence, (1) 1s
. the line of regression of Y on X and (2) is the
⇒ y - 127.22 = -2.00 (X - n 22> line of regression of X on Y.
⇒Y-127.22 =-2X + 154.44
⇒ y = 2X + 154.44 + 127.22 Re-writing (1), we get, 5Y = 4X + 30 ⇒ Y =4 X+6 ⇒
⇒ y = 281.66 - ·2X
4
5
bYX=5
S.1.8. Relationship between Correlation
& Regression Coefficients (2), we get,
Re-writing 20X = 9Y + 107 ⇒
The correlation coefficient is the Geometric mean 9 107 . 9
between the regressjon coefficjents i.e., X=20 Y+W ⇒ bxY= 20
2
⇒
⇒
•r =byxxbxy r=Jbyxxbxy ..... (1) 2
r =bvx ·hxy=ix-2_=0.36
5 20
r=±v'0.36=±0.6
Let us suppose that the two lines of regression are:
aJX + bJ Y + CJ = 0 ..... (2)
a2X + b2 Y + Ci = 0 ..... (3) Since both the regression coefficients are positive, r mu 5l
be positive. Hence, we take r = rxY = 0.6
Let us suppose that equation (2) is the line of regression of y
on X and equation (3) is the line of regression of X on Y. We are given, crx = 3
From equation (2),
iii)
aJX + b1 y + CJ =0 or b1 y =-a1X - C1 We have, bYX =r~ ⇒ cry bYX · 0-x _ 4 x2=4
crx r 5 0.6
Example l6: In a Partially destroyed laboratory recortl of~
ananceofofCorrelations
analysis
V x == 9 data, the following results are legibl ·
 - ssion (Chapter 5)
~ef•• 101
dft:ssion equations: 4x -Sy
~ Ci,· 22x + 33::: 0 Substituting the values, we get
15
·nd out: - Y::: 108 Oy 2.4 324
ft ]VleaD values of x and y
,) d dd . ·
0.8 = 0.74x
3 ⇒ Oy = 0.? 4 = ·
ii) stan ar_ eviation of y
Hence, x=20.7, y=23.16, r=+0.74, <Jy =3.24
,. ') coefficient of correlation b
,u etween x and Y
Example 17: The two regression lines of the variables X
sotution:. . and y are: ·
i) fhe given equations are:
16x -6y + 60 = 0 and 30x -9y- 150 = 0
4x - 5y + 33 = 0 or 4x _ 5 _
Y--33 Find out means of x and y and Correlation Coefficient
22x -15y::: 108 ... .. (I)
2 between two variables.
Since the mean of x and y Ii · · · ·· ( )
4 x- 5 Y= - 33 e on the regression line. So, Solution: Given, First regression equation,
16x - 6y + 60 = 0
22x -15Y = 108 .. ... (3) ⇒ 2(8x - 3y + 30) = 0 ⇒ 8x - 3y + 30 = 0
····· (4) ⇒ 8x-3y=-30 ..... (1)
Multiply 3 in equation (3) Now, Second regression equation,
12x -15Y=- 99 '
..... (5) 30x-9y-150=0
⇒ 3(10x-3y-50)=0 ⇒ 10x-3y-50=0
Subtract equation (5) from (4) ⇒ lOx - 3y = 50 ..... (2)
10x=201 •
Subtracting equation (2) from (1), ⇒ '._ 2x = - 80 ⇒ x = 40
207
⇒
x= 10 =207
. Putting x = 40 in equation (1), we have,
350
8(40)-3y =-30 ⇒ -3y =-30-320 ⇒ y=3
Putting the value'!! X in equation ( 3),
4 X 20 . 7 - 5 y = - 33 350
:. x=40and y=
⇒ 82.8-5Y=- 33
3
⇒ -5y=-33-82.8 :. X = 40, Y = 116.6667
⇒ Y=23.16 Suppose 16x - 6y + 60 = 0 is regression equation of y on x,
⇒ 8x + (-3y) + 30 = 0 ⇒ - 3y = - 8x + (-30)
Hence means are: x=207; y=2116

ii) Calculation of Coefficient of Correlation (rxy) ·

⇒ y=(- ~ 3}+(- ~~) ⇒ y=2.6667x+ 10
Let us assume that the first equation be regression byx = 2.6667
~~oo~yoox '
Suppose 30x-9y-150 = 0 is regression equation ofx on y,
4x-5y + 33 = 0
33 4 ⇒ IOx + (- 3y) + (- 50) = 0 ⇒ lOx = 3y · + 50
⇒ 5y=4x +33 ⇒ y=-+-x 3 50
5 5 ⇒ X = 10 y + fO ⇒ X = 0.3y + 5
⇒ y=6.6+0.8x
byx = 0.8 bxy = 0.3

Assuming second equation as regression equation of x

r = ✓ bxy· byx =✓2.6667 X 0.3 =.Jo.8 = 0.8944
ony: Example 18: Find the mean of X and Y variables· and the
22x- 15y = 108 coefficient of correlation between them form the following
⇒ 22x= 108 + 15y two regression equations:
⇒ x= IOS +~y ⇒ X = 4.9 + 0.682y 2Y - X =50, 3Y - 2X = 10
22 22 Solution:
bxy = +0.682 The given equations are:
Now r = lb . b
' V xy yx
=✓o.682X0.8 =-t-0.74 2Y-X=50 ..... (1)
3Y-2X= 10 .. ... (2)

iii) Calculation of CJy Since the mean of X and Y lie on the regression line. So,
Given, variance of x=<fx =9⇒crx =3 2Y-X=50 ... .'. (3)
3Y-2X=10 .... . (4)
byx = 0.8, r = +0.74
Multiply 2 in equation (1),
<Jy 4Y-2X=100 ..... (5)
We know, byx =r·-
<J"
 . semester (Quantitative Tec:hnlqut,) Qc"Ill,.'
..
~=A First
lY= calculate: i) Correlation ij\- ·1
Jtegte!
the following da;X and S.D. X = 3 ' 19,l)_Iii,
l02 (Unir I) 3) From y y = O. r
X = 0.854 - 3 06] , f'ill
. (4) from (5 ). -087,S,D , ofY-. i)
Subtracrequauon ii)
[Ans: r - . . data obtain the two re cssio11 U
Y=90 the foUowin IO 4 8 n~:
- . c uation (3). X = J30 4) From X 6 2 5 8 7
Puning the value~ Y m,; _ 5[ = 50 ⇒ y 9 J....;l~l'--'-_ _ _:-:-
2x90-X=50 180
y + o.65x = 11.9] JZ)
I Jy = 16 .4 •
Hence means are, _ [Ans: x + ' . data obtain the two re re?!ibn lt,.
X = 130 =90 . be regression 5) From the foUowlll 5 3 2 I 1 7 3 - -1:

Let us assume that the

equation of Y on X.
Y
first equation
X
X ! l O O 1 2 l
. , , when x = IO 13)
y- . Predict y 25
2Y-X=50 ⇒ - --25
2 1)
'")
11
. .
Predict
, , when Y = ·
x ·
_ 2 09 x + 0.28y = 3.56, (i) O.]0, (ii) ,
I : O.3x + Y - · • '-861
:. by,: = 2 [Ans . n coefficients for the foll owing data;
. equ ation of X3 on Y. 6) Calculate re r;ss1~ 6 8 10 ' 12 14 ' .
. will be regresston
Second equauon X- - y - 5 yX 4 2 5 10 4 11 12
3Y - 2X = JO ⇒ 2X = 3Y - IO ⇒ - 2 14)

3 [ADS: bxy -- 0 ·85 ' bY" = 0.73]

b,y = 2 7) You are ve
. n the followin data:
X y
Variable . 47 96
Now,
,---:--_
/H I{
2x..!.. = 2 = Jo.75 = 0.866
Mean
Variance 64 8l
r = Jb,y . by, - 2 2 4 ffi . y is
t of correlation between X and
0
Coe c~en the equations of regression lines. Calculatt-ll. 15)
wen ~e50
DhetenruX - , and X when Y = 88, ~
.2. EXERCISE
I .. [Ans: Y::0.405X + 76.965, X=0.32 Y + 16.28, X "'44 4i
Y=97.215]
5.2.1. Short Answer Type_Questions
sion coefficients. . .
I) Write the properties of regres_
S) Derive regression lines for the following data:
2
l) Whal do you mean by regression· lines and equatJons? . -DC= 30, DC = 190, :EXY = 192, LY= 30, tY2 :: 190,n,i 16)

3) Write the difference between the correlation and regression [Ans: X =- 1.2 + l.2Y, Y =- 1.2 + l.2X]
analysis.
4) Write the application of regression analysis. 9) You are given the following information about advellisiit
expenditures and the sales: ·

5.2.2. Long Answer Type Que~tions . Advertisement Sales

1) Write the definition of regression. What 1s regression tin Lac) (~ in Lac)
coefficient? Arithmetic Mean
10 90
2) What do you mean by regression lines and regression Standard Deviation
equations? 3 12
3) What are the applications of regression analysis? The coefficient of correlation is 0.8.
4) Why are there two regression lines? When do we use one in i) Obtain two regression equations.
preference to the other?
ii) Find the likely sales when the advertiseme nt budgeti
5) Explain the relationship between correlation and regression fl5 lac.
coefficients.
6) What do you understand by the tenn 'regression'? With the iii) What should be the advertisement budget if llx
help ?f fe~ examp!e~ illustrate how regression analysis company wants to attain sales target fl20 lac?
helps m busmess dec1S1on making.
[Ans: i) Y =3.2X + 58, ii) Y =106 Iac, iii) X == 16 lac]
5.2.3. Practical Questions 10) Write regression equations of X on
following data:
y and of Yon X forti'/.
1) From the following data obtain the two regression equations:
Sales91 97 l08 121 67 124 51 73 111 57 X 45 48 50 55 65 70 75 72 80 85
5
Purchase 71 75 69 97 70 91 39 61 80 47 y 2 30 35 30 40 50 45 55 60 65
=
[Ans: x -5.2 + l.36y; y 15.I + 0.6IxJ = [Ans: X = O.OIX _ 21.64S]
2) Two regression lines a.re 3X + 2Y _ _
ll) The information about sales turnover and a<lvert~~!
= 0. Find the mean value and the
26 0
co - . and 6X + 31 '! - ex enses iven below:
the variance of Xis 25 then find S.D. :;~~on coefficient. If
Mean S.D.
[Ans: Mean ofX =4, mean ofy =7, r::: 0 Advertisement Expenditure (\'lac)
10 3
·5, s.n. ofy::: 15] Sales turnover (tcrore)
90 12
Coefficient of correlation
 ,,ion (Chapter 5)
11eFfC' ·

f ind: . 103
i) Tw_o regrelilc~s1oln equations.
•') Esnmate e y sales turn' . tof
11 17) P-md the means of X and Y variables and the coeffi cien .
budget is 15 ~ac. _over "-'hen the
. advertise correlation between them from the following two regression
[J\ns: 1) Y = 58 +3.2X, ii) 106 l..acJ ment equations:
find out the regression eq . 4X-5Y+33 =0
J2) frOW the followin data; llation by least 20X-9Y-107 = 0
X 1 3 4 6 sqllare lllethod
8 9 II [Ans: 13, 17, r =0.6]
I 244 5
. . .. . ' . 7 8 18) The following table gives the aptitude test scores and
[ADS! X=-0.5 + f.5 y and Y-o· 9
productivity indices of IO workers at random:
coeffi - . ·546+063· XJ
calculate the regression
JJ) regression for the followm ,1A.~ient and obtain the lin Apdtude score 60 62 65 70 72 48 _53 73 65 82
X I 2 3 - esof
4 5 Productivity Index 68 60 62 80 85 40 52 62 60 81
6
~ Y~9__,_-'--8-L....!I~OJ_l~2u_!JLJ 7
·1 ' 13 ' 14
[Ans: b,y =0.929, byx= 0.929· X=O Estimate:
.
929y -6.219, Y =0.929X i) The test score of a worker whose productivity
+ 7.284] '
index is 75.
re8I'CSsion , , - . · · ·: ·
14) Calculate the two 0 ii) The productivity index of a worker whose test
X from the data _givcn_ beio~~- fX-;~ n.YandYon • scor~ is 92.
actual means of X and Y. · · g_ '!_e_y11~tions from a
Price f 10 -12 , _13•. , [Ans: i) X7s = 70.96, Y92 = 96.536)
Amount Demanded 40 38 · l2 • 16 · 15
43 45 37 43 19) Obtain the line of regression of Y on X by least square
Estimate the likely demand ~hen th ;: :- -. method for the following data:
- - e Rnce~1s ~2.0.
s
I ~
[Ans:X= -0.12Y+l7.92 Y=-025X ,-,, A· · 4
' · .,. -.25, Y=:39.25] I; I; I; 4 6
15) Obtain regression equation of y on X and estimate y when
X = 55 from the followin . Also obtain an estimate of Y when X =2.
X 40 50 38 60 65 50 35
Y 38 60 55 70 60 48 30 [Ans: Y =1.3 +o.9 X; 3.1)

[Ans: Y = 0.942X + 6.08, Y= 0.942(55) + 6.08 = 57.89] 20) Find the regression of Y on X and X on Y by the least
square method for the following data:
16) Fmd the means of X and Y variables and the coefficientw f
corre~ation between them from the following two ~ gres~ion X
equations: y
2Y- X-50=0
3Y-2X-l'O = 0. J'!.. Also find coefficient of correlation.

[Ans: 130, 90, r =0.866] [Ans: Y =0.667 +l .5X; X = -0.357 + 0.643 Y; r =0.982]