So一 Form givenby

Regression
RegressionSo sales substi Thus 00一0
Business乛
the Mean
University
Statistics 冖 …
冖
Correlationcoeffic :0
andCoefficient
Standard Standard

The
Mean
~ Standard
E Calcula 0
冖e
a20 pte 冖
冖 h
sston

0 marks
he Business
Marks
… X Y—V
adver
丶
0 s冖
imate a e uting 冖
0
0 mo
par
Marks

一 、 … in x | lakhs. 0 mo Analysts

0 R line tine b ·
冖 X
、
We de
冖
Statistics•
X The
冖
secured e icular
s
一
Deviation

X 冖 = isemen
Deviation

0 冖 冖Ⅱ
o n
ofcorrelation
冖
thethe
一
in B.Com/
00ng0 0.2 " Ⅱ " are likely
40
in

、
0 Business 0 following Y 0
equ
Statistics
、 、 0
、 x 冖bwX0 byxon 一
y
Advanced

regression
10,given adver …
e in
、 firmproduc
Then

冖
byon 一 (Y 冖
20expenditure
X—R)
It
16X 冖
Eq•(l)' 、
一 p
、
oduc
ofMarks
(Hons•)
on
ofMarks

hestudent
a lines
|8 | … X一 0.8 the isement
冖
m da冖 + 0
between

一 冖
we

and a 20 一
Accounts

= or, or, × 90 冖 Y on
equation ion
we
following
s a
、
一
on
are
、
in inAdvanced

given
…
:A in 0 、
Business

year 一
6 required 3 12 expenditure Adv.
COrresponding
0 冖 ge 00
xwhich
、
Bucu x— Y given
、0P0n
he

~ 一 lakhs• Y 0
,、 、 a
Marks
of
Expenditure
information

ne冖
…
Examina 一 90 3 3
n00 一 0 66P or
(€lakhs)

冖 0 ,2 一 Y 3
he =
10
regarding
、 0 ()•2 ,3.2 in
on
Accou

、0XP00
ofAdvancedAccounts
冖 and、
Accoun ~
a

S a = required
X. Y
| 一
500 冖
冖 冖
、
s

冖
一
、 冖
、
Its 冖
ion tain
()¯ 一 X b
一
2…
0
the e n
0
冖 ed | by
~

冖 一
Deth
0 0 a 90 一 and 冖 C
expenditure
0
、marks
0.8x 、
rainfall
、
冖 sales 0 ·
00丶 0
a 一 一
— =>
冖
r at冖ain
0
·Foundation,
0.8 、
in
Y 丶40.in
Un
par
乛
、
0.8 a
冖
=
OY
arge
000 Advanced
冖
icular X Y= 0
乛
0
·B.Com/ Ⅱ on 40 5 冖
100

冖
冖
0 0
乛O•2 3.2X 0 sales cms
advertising 冖
x 0
0 =
year
一
20Y¯
·8
× NO
&lakhs) Sales
| 0m0
m(P)Accoun 35 12 3 12 90 is 30
」
we
+ target
10 30

in
8 58
lakhs 660 一
0 7

201q
、
8
1999
、
…
Delhi
0.2 0
、
n is 一 Of
andkg. 9n3
 Business
7 and r
- 1.143

and
RegressionEquations
: RegressionEquationof Y
Regression Equation of X on V is on R .

. - (Y -35) or, Y -35- 0.56 (X-3())

Y -35= 0.56X- 16.8
X- 10 Y - 0.56X+ 18.2
(XOif the marks in Business Statistic
Te find the expected mavks in Advanced Accounts
are 40, use regression line of X on Y.Thus substituting Y = 40 in (1),we 3
35.72. get
x - 1.143 40-10 -
Evtnpte 12. From the followingdata obtain the two regression equations and estimate
Nalueof X when Yis 130 and estimate the value of Ywhen X is 30.
39 33 30 31 32 36 41 49 46 43
132 134 138 129 136 131 132 135 128 125
[Delhi Univ. B.Com. (P)
20121
Solution. Calculations For Regression Lines

39 132
33 134 25 -10
30 138 64 36
31 129 49
21
32 136 36 16 —24
36 131
41 132
49 135 11 121 33
46 128 64 16 —32
43 125 25 49 —35

EX
= 380 = 1320 = 398 = 140 = -93

380 1320
= 38 and V = = 132
10 10
-93 —93=
—0.23
140 = 0.66 and b
398
 ssionAnalysis 9.15
Equation of X on Y Regression Equation of Y on X
Regression
Y -V = byx(x-R)
-0.66 (Y- 132) or, Y- 132 = -0.23 (X- 38)
of,
= -0.66Y+ 87.12 or, Y- 132 = -0.23X + 8.74
X-38
or,
x = -0.66Y+ 125.12 ...(1) Y = -0.23X+ 140.74

Toestimate the value of X when Y= 130, we use regression equation of X on Y.Thus

substituting Y= 13()in Eq. (1), we get
x = -0.66 x 130 + 125.12=-85.8 + 125.12= 39.32
Toestimate the value of Ywhen X = 30, we use regression equation of Yon X. Thus
substituting X = 30 in Eq. (2), we get
Y = -0.23 x 30 +140.74 = -6.9 + 140.74 = 133.84.

Example 13. The followingdata give the aptitude test scores and productivity indices
at random :
of 10 workers selected
Aptitude scores (X)
60 62 65 70 72 48 53 73 65 82
Productivity index(Y) : 68 60 62 80 85 40 52 62 60 81
calculate the two regression equations and estimate the productivity index of a worker
whose test score is 92.

Solution. Calculations For Regression Equations

Aptitude Productivity
Score Index
x=x-R xy

68 25 -15
60 15
60 25
62
65 62
15 25 225 75
70 80 400 140
20 49
72 85 425
-17 -25 289 625
48 40
-13 169 156
53 52 -12 -24
64
73 62 25
65 60 272
17 16 289 256
82 81

= 894 = 1752 = 1044

= 650 = 650

650 —65 650 = 65

10
44
1.168 and 17
894
 Analysis 9.25
ssiotl
_ 46
7
- 58 +6.57 - 64.57
n
_ 58+2
_ 58 + 1.29 59.29
7
Equation of Yon X is : Regression Equation of X on Y is :
Begression

0.37 (X- 64.57) 64.57 = 2.20 (Y- 59.29)

Y- 59.29 = or, X-

Y- 59.29 =
0.37X- 23.891 i.e., X- 64.57 = 2.20Y- 130.44
Y = 0.37X+ 35.399 x = 2.20Y- 65.87
-.2

lit)
Coefficientof Correlation is given by
b xy x b yx 0.814 0.902.

EX
te 25. You are given the followinginformation regarding a distribution :
N = 5, R = 10, V = 20, - 100,

-10)2 = 160, = 80.

Findthe two regression coefficients and hence the coefficient of correlation.
[Delhi Univ. B.Com. (H)2002]
Solution. If we let U = X—4 and V = Y—10, then we are given :
EU2 = 100, E v 2 = 160 = 80
d peh
Also, 4 EU = E(x-4) = EX-4N = NR-4N = 5x 10-4x5 = 30

= E Y -ION = NV-ION = 50
and = E(Y-IO)
Since the regression coefficients are independent of change of origin, therefore,

5 x 80-30 x 50
b b

400-1500 -1100 11
500 —900 = —400

and b

400 -1500 -1100 11

800 - 2500 -1700 17
Coefficientof correlation (7)is given by
11 11 121
b xy x b yx 1.334.
17 68
But r can't exceed I. Hence the given data are inconsistent.

xampZe26. In order to find the correlation between two variables X and Yfrom 12
pairs of observations, the following calculations were made •
E x = 30, E x 2 = 670, E Y = 5, EY 2 = 285, E xy = 344
9.20 Business
statistics
t hat the pair (X- 11, Y- 4) was
On subsequent svnfication it was discovered making
the being (X- 10, Y- 14).After
find:
the tvgvesston coefficients,
the tvgtvssion equations, and
[Dethi Univ. B.Com.(R)
(iit) the eovtvlation coefficient. 20141
Solution. We have
Cormcted Y X 30-11+ 10 - 29
= 649
Corrected X2 = 670-11 2 + 102
Corrected YY 5—4+ 14 = 15
- 285-42 + 142 = 465
Corrected Y2
344—11 x 4+ 14 = 440
Corrected Y XY =
Corrected X = 29
h) Corrected R = 2.42
n
Corrected Y
= 15
Corrected = 1.25
n
(i)Regression coefficientof X on Y :
5280 - 435 4845
b 5580-225
12 5355 - 0.905
Regression coefficient of Y on X :

x 15 4845 4845
b 6947 0.697
12 x 7788-841
(it) Regression Equations
Regression Equation of X on Y is : Regression Equation of Y on X is:

x -R = bxy(Y-V) Y-V = byx(x-R)

or, X-2.42 = 0.905 (Y- 1.25) or, Y- 1.25 = 0.697 (X- 2.42)
i.e., X-2.42= 0.905Y- 1.13 Y- 1.25 = 0.697X- 1.69
X= 0.905Y+ 1.29 Y = 0.697X-0.44
(iii) The correlation coefficient (r)is given by
0.905 x 0.697 = + 0.79.
ample 27. For 25 observations on price (X)and supply (Y),the following data were
obtained :

R = 5, V = 4, E x 2 =680, E Y 2 = 480, E xy = 508

It was later on discovered that two pairs of observations (X, Y)were noted as (8, 10)
and (5, 8) while correct values were (10, 12) and (3, 6) respectively. Determine the
corrected values and find :
 Analysis
9.27
qsion
two regression coefficients,
(l) the
correlation coefficient, and
the
when supply
-— 80 units and the supplywhen price= 70.
the price
[Delhi Univ. B.Com. (H) 1996]

sottitiort. n
(Incorrect)EX = 25 x 125
(Incorrect) E Y = 25 x 100
and Corrected x 125-8-5+ 10+3 - 125
Corrected Y 100- 12+6 = 100
Corrected X2 680 —(8)2 —(5)2 + (10)2 + (3)2
680 - 64 —25 + 100 + 9 = 700
Corrected E Y2 480 —(10)2 _ (8)2 + (12)2 + (6)2
480 - 100-64 + 144 + 36 —496
Corrected XY 508-8x 10-5 x 8+10 x 12+3 x 6
508-80-40 + 120 + 18 = 526
Corrected X X 125
CorrectedR n 25
Corrected Y 100
Corrected V n 25
(i)Computation of Regression Coefficients :
25 x 526-125 x 100 526-500 26 0.271
496 —400 96
25 x (100) 2
25 x 526-125 x 100 526 - 500 26 0.347
b 700-625 75
25
(ii)Thecorrelation coefficient (r) is given by
= + 0.3067

(iii) Regression Equations

Regression Equation of Y on X is :
RegressionEquation of X on Y is :
bw(X-R)
x -R = bxy(Y -V) Y -V =
or, Y-4= 0.347 (X- 5)
or, X-5= 0.271 (Y- 4)
i.e., X-5= 0.271Y- 1.084 i.e., Y-4 = 0.347X- 1.735

X = 3.916 + 0.271 Y . (1) Y = 2.265 + 0.347X (2)

TO estimatethe price when supply = 80 units, we substitute Y = 80 in Eq (I) .

= 3.916 + 0.271 x 80 - 25.596
Toestimatethe supply when price = 70, we substitute X = 7Öin Eq (2) :
70 - 2.265 + 0.347 x 70 26.555.
 ssion Analysis 9.31
Regre
our assumption is correct. However,
byx.bxy 1
5'
sitive sign is being rejected, because r, byxand b all have the same signs.
tile po
standard deviation of Y,if
(iii)Calculating = 25
= 25
3 3 3
b 15
Also,
33. If the two regression lines are 3 Y+ 9X= 46 and 3X+ 12 Y= 19, determine
which one of these is the regression line of Y on X and which one is that of X on Y.
Alsofind the means, correlation coefficientand ratio of the variance of X and Y.
[Delhi Univ. B.Com. (H)2006 (SOL)]
solution. The equations of two regression lines are :
3Y+ = 46 (1) and 3X+ 12Y= 19
Let us artibrarily assume that Eq. (1) is the line of regression of Xon Y and consequently
the line of regression of Yon X.
Eq. (2) is
(or X in terms of Y, we obtain
Solving Eq (1)
1 46 1
3 9
Similarly,solving Eq. (2) for Yin terms of X, we obtain
1 19 1
4 12
F = byx. bxy 1 1 1
< I, which is true
Hence our assumption is correct — that is, the equation 3 Y + 9X = 46 is the line of
of Yon
regression of X on Y, and the equation 3X + 12 Y = 19 is the line of regression
x.
Thecorrelationcoefficient (r) is given by
bxy x byx - -0.289,
the negative sign is being taken, because r,b all have the same
and b.r,x signs.

Calculation of means of X and Y

To find means Y—series, we solve Eqs. (1)and (2) simultaneously
of X and for X and Y.
Multiplyingboth sides of Eqs (1)by 4, we obtain
36X+ 12Y = 184 (3)
Subtracting(3)from (2),we get
- 33X = - 165
Substituting X = 5 in (I)we get
 Business
9.32 Statistics
1

Thus, the means of X and Y—series are : R 5, Y

and Y
Calculation of ratio of the variance of X
variance of X and Y, we make use of the following formu la.
To find the ratio of the

1/9 12 or 4 : 3.
2 1/12
4 are the lines of regression of X on
ample 34. Given X = 4 Y + 5 and Y = RX+
cannot exceed —. If
k=
Yon X respectively. If k is positive, prove that it 16'find
the
and the coefficient of correlation between them.
means of two variables [Delhi Univ. B.Com. (R) 20061
by
Solution. The regression line of X on Y is given
X = 4Y+5 bw
The regression line of Yon X is given by
Y = Wc +4 byx
r2=b xb = 4k
1
r 2 S I (always) 4k < I
1
If k = —
16'
en the regression line of Yon X is :
1 1
16 16
1 1 1

16 2'
the positivesign is being taken, because r, b and b all have the same signs.
Since both the lines of regression pass through the mean values (R, V) , therefore

1
. (1) and Y
16
Solving Eqs. (1)and (2)simultaneously for and V ,we obtain R = 28 and V = 5.75.

9.6 STANDARD ERROR OF ESTIMATE

As discussed earlier, the primary use of the regression equation is to estimate values
of the dependent variable from knownvalues of the independent variable. However,
the natural question is : "Howreliableis the estimate?" Obviously, an important
factor in etermining reliabiliWis the closeness of the relationship between the
variables. or example, let's look at figure showing two scatter diagrams, where
assume that both scatter diagrams have the same scale for the variables and the
same regression line.
RN'ession Analysis
35. Find the line
of regression
9.35
X from the
followingdata :
calculate the standard 12
error 14
TABLE 9.1. of Yon X. In 12
Calculation for
Regress: Lin
2 4 4
16
4 3 16 8
9
5 12 1.1784 1.3886
6 36 4.2144
25 - 1.2144 I.4748
8 10 30
64 100 5.6072
4 80 -0.6072 0.3687
10 100 7.0000
16 40 3.0000 9.0000
12 11 144 8.3928
121 —4.3928 19.2967
12 132
14 196 144 9.7856 1.2144 I.4748
168
11.1784
0.8216 0.6750
= 56 49 = 560 = 431 E(Y-Yc) -Yc)2
= 470 49
Note that E Y and E YC are equal. = 33.6786
This must always be
true if E (Y —Yc)

and V Y 49
b
— 7X ¯ 56 x
7 x 560 - (56) 2
3290 - 2744 546
Regression equation of Y on X
3920-3136 784 = 0.6964
is given by
Y -V = byx(x-R)
or, Y -7 = 0.6964 (X- 8)
Y = 1.4288 + 0.6964X
Tocalculate the standard error of estimate
(of Yon X),we must compute a value for
Ycfor each value of X by substituting each X value into the regression equation (I).
We must then compute the difference between these Yc values
and the corresponding
values of Y.These values are shown in Table9.1.
The standard error of estimate of Y on X is calculated as follows :

E (Y -Yc)2 33.6786
7
= = 2.1934

Alternative Methodfor Calculating Syx:

The standard error of estimate S can also be calculated using the followingformula:
EY2-aEY-bEXY
n
9.36 Business
Statistics
-0.6964 (470) a- 1.4288, b - 0.6964)
7

70.01 -327.308
7
- 2.1935 (the differencein this case is due to rounding)
Example $6. In fitting of a regtvssion of Yon X to a bivariate distribution consisting
of 9 obsenations. the explainedand unexplained variations were computed as 24
and 36 tvspeetivebL Find (t) the coefficient of determination, and (it) the standard
ermr of estimate of Yon X. [Delhi Univ. B.Com. (R) 2007
(SOL))
Solution. We are given :
n 9, explained variation =24, and unexplained variation = 36
Total variation -Explained variation + Unexplained variation = 24 + 36 60
Coefficient of determination (r 2) is given by
Explained variation = — 24
Total variation 60
= 0.4
(it) Standard error of estimate of Yon X is given by
Unexplained Variation 36
sYX
Exarnple 37. Given that: n = 10, Total variation = E (Y —V) 2 = 38.92 and
Unexplained
Variation = ¯YC) = 19.7. Find : (t)coefficient of determination, (it)
coefficient of
correlation, and (iit)standard error of estimate.
Solution.We have
Totalvariation = Explained variation + Unexplained variation

38.92 = E(Yc -V + 19.7

= 38.92-19.7 = 19.22 i.e., Explained variation = 19.22

(O Coeffcient of determination (P) is given by
Explained Variation 19.22
0.49
Total Variation 38.92 =
(it) Coefficient of correlation is given by

= 0.7
(iit) Standard error of estimate is
given by

s Unexplained Variation 19.7

10 1.97
Regressi Analysis
pie 3& Given the following
data • 9.37

10
calculate: (t) the regression
explained variation, and (it,) coefficient of Yon
the standard X, (it)the
error total variation in Y,
Computation of estimate the
of Regression of Yon X.
Coefficient
x-i Y -V
(Y- Yc)2
= 4.29 + 0.53X

7.47
1.53 2.3409
6.41 - 0.41 0.1681
8.00 1.00 1.oooo
6.41 - 1.41 1.9881
16 8.53 0.47 0.2209
10
5.88 0.12 0.0144
9.59 -1.59 2.5281
11 7.47 -0.47 0.2209
8.53 2.47 6.1009
8.00 0.00 0.0000
12 11 9.06 -2.06 4.2436
15 10.65 0.35 0.1225
= 84 = 96
= 76 = 40 = 40
- 18.9484
x 84
12 and = 96
(i)Regression coefficient of Y on X 12
:
Exy 40
76 - 0.53
Regression equation of Y on X :

Y -V = byx(x-R) or, Y -8 = 0.53 (X- 7)

Y = 4.29 + 0.53X

(ii)
Total variation in Y Y — 40
(iii) Computation of Explained variation. To compute explained variation, we
first
compute unexplained variation:
Unexplained variation = 2
¯ Yc) = 18.9484
Explained variation = Total variation— Unexplained variation
- 40- 18.9484 = 21.0516
 Business
9.3S
on X :
Standard error Ofestimate of

(10 the following

(N) (Into were
$0.
obtAtned -5506, y XY
of h' on A'and estitnate the supply when the price in 16
the line of est innate. [Delhi Univ. D.Com. (R)20201
untts, and find the standatd
of Yon X is :
Solutions Regression eoqmcient

-cyx) 2288 -(130)?

34070 -28600 6070 - 1.015
"SS6Ti6500 '"' 5980
Vx 130 -13 and 10 22
10
Regression equation of Yon X is :
or, Y-22 - 1.015 (X- 13)

Y - S.805 + I.Q15X
obtained by substituting X = 16 in
price is 16 units, the estimated supply is
8.805 + 1.015 (16) = 25.045
by
The standard error of estimate (of Yon X) is given

n
5506-8.805 (220) -1.015 (3467)
(a = 8.805 and 1.015)
10
5506-1937.1-3519 49.9
4.99
10 10
Example 40. The followingdata relate to the scores obtained by 9 salesmenofa
company in an intellignece test and their weekly sales performance:
Salesmen
Y Test Scores 50 60 50 60 80 50 80 40 20
U Weekly Sales (in '0000 8) 30 60 40 50 60 30 70 50 60
(t) Obtain the regression equation oedellfgen test scores f the salesmen.
(it) Jf the intelligence score of a salesman is , what would be his expected weekly
sales?
(iit) Calculate the standard error of sales on intelligence test scores.
y [Delhi Univ. B.Com. (H)20131