0% found this document useful (0 votes)
50 views20 pagesRegression 9
Uploaded by
Manish SandilyaCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
50 views20 pagesRegression 9
Uploaded by
Manish SandilyaCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 20
5 GRESSION
The term ‘regression’ literally means “stepping back towards the average’.
Regression analysis is a mathematical measure of the average relationship
between two or more variables in terms of the original units of the data.
537“Line of Regression
The line of regression is the line which gives the best estimate to the value of
one variable for any specific value of the other variable. Thus the line of
regression is the line of best fit and is obtained by the principle of least squares.
The regression equation of Y on X is obtained on minimizing the sum of the
Squares of the errors parallel to Y axis while the regression equation of X on
"is obtained on minimizing the sum of the squares of the errors parallel to X
ais,
Ifthe variables in a bivariate distribution are related, then the points in the
Scatter diagram (which is a diagrammatic representation of bivariate data) will
‘hister round some curve called the curve of regression. If the curve is a
Sttight line, it is called the line of regression and said to be linear regression
between the Variables, otherwise regression is said to be curvilinear. 410 & Probability and Random Processes
5.522 Equations of Lines of Regression
‘The regression linc of y on x 1
-y=p—(x-%
y-Fap>(x-3)
which is used to predict or estimate the value of y for ANY given
‘The coefficient of x in the regression line of y on x is tye
coefficient of y on x and is denoted by ty
aja
b,,
The regression line of x on y is
pZ(y-j)
a (P3
which is used to predict or estimate the value of x for any given y
‘The coefficient of y’ in the regression line of x on }’ 18 the teen
coefficient of x on y and is denoted by
The correlation coefficient p in terms of the regression coefficients is
Dat by Pye
5.543 Properties of Correlation and Regression Coefficients
Correlation coefficient is the geometric mean between regression
coefficients, i.c.
pt yby by,
The correlation coefficient cannot numerically exceed the arithmetic me2*
between regression coefficient.
If one of the regression coefficients is greater than unity, the other mst
be less than unity numerically.
Regression coefficients are independent of the origin, but not scale
The correlation coefficient and the two regression coefficients have
|
|
xv
s
. m
The regression coefficients are obtained by the following expressions
case of discrete values of X and ¥: Twordimensional Random Variables 4 411
Oy _ nexy= (EH) (Ey) _ Ue 5)(y—Fy
bye PG. EEX ene
oy _ MEAY= (ZH) (By) _ Ux-F)(y~F)
-e ee By)
oy nby? = (Ey) Ly~ 5
“angle petween the Regression Lines
554
jeetneen te rogression lines is given by
le
‘he ang!
sate:
g Whon r= #1, fan 0=0, ic. O= 0 or x. In this case, the two lines
ee x F
of regression coincide. When o=55 the lines of regression are
perpendicular,
{iy Whenever the two lines intersect, there are two angles between them,
one acute angle and the other obtuse angle.
fi) Both regression lines pass through the point (x, ¥) where ¥ = mean
ofx and ¥ = mean of y.
(i) Regression curve of ¥ on X is y = E(YIN = x)
and regression curve of X on ¥ is x = E(X/Y = y).
EXAMPLE F vein the correlation coefficient for the following heights
(ininches) of fathers X and their sons Y. {AU June 06, November 07]
X | 65 | 66 67 | 67 | 68 | 69 | 70 | 72
Y 67 | 68 65 | 68 72 | 72 | 69 | 71
Solution Method 1:
x ¥ XY. me 7
| 6 67 4355 4225 4489
ie 68 4488 4356 4624
i 65 4355 4489 4225
Hes 68 4556 4489 4624
6 “72 4896 4624 5184
0 72 4968 4761 5184
Lin 69 4830 4900 4761
lino aN S112 5184 5041
S284 | nay = 552 [ex = 37560 | £.x2 = 37028 | 21? = 38132 | 412 & Probability and Random Processes
Now,
RY = 68 x 69 = 4692
loy2_ 722 37028 604 =
a puiee ax? f= 2131
<2 _ [38132
a Pay? 7? = =z 7 4761 = 2.345
= 4—__ = =>—_—____—_- = 0,
Oy 2I2ixa2s45 6030
Note: Correlation coefficient is independent of change of origin ang Seal,
ie. :
n(X, Y) = (U,V)
where
u= X-a
ok
¥-b
id a
ant Ve K
Here a and b are some arbitrary constants and usually the mid-values of the
given data X and Y respectively. pLE an Find the coefficient of correlation between industrial
woe and export using the following data: [AU May 706]
Production (x)| 55_| 56 | 58 | 59 | 60 | o | 62
Export (9) Sil eo OHe |e 32d |e om 44a ease lead
Solution
X=x—58| Y=y-40 AY. x2 3
3 -5 15 9 25
-2 -2 4 4 4
0 3 0 0 9
1 cl -l 1 1
2 4 8 4 16
2 3 6 4 9
4 4 16 16 16
zxy=4 zY=0 EXY = 48 | EX? = 38 | EY? = 80
x = 4 Losing
no7
pe ZX 0g
n OF;
Cov(x, Y) = EXY _ gp 248 _ 96.857
n lity and Random Processes
2_ Bs
= [= - (0.5714)
7 (0.5714) = 225%
sy? z7_ [go
y= art a7 = | o=aae
_ Cov(X¥)_ 6.857
= Fg, 2258x338
414 @ Probabl
0.898
EXAMPLE 5/11 Find the correlation coefficient for th folloyi
ng dy
(x [10 [44 [1s [22 | 26 [30
[> [as [2 [2% [6 | 30 [36
Solution
cl YY TA
10 | 18 3 3 F
4} 12 -2 4 a
18 | 24 = 0 ;
22 6 0 0 0
26 | 30 1 1 1 ;
30_| 36 2 2 Fi ‘
mY=-3 xy=-3 | 2av=12 zits 9 [ern]
Cov(X, Y) = ——— X¥
12 7
= FON AON=F
Sx? =) fd
Oy = —- X? = f= - (0.5) =1.708
s =e eS)
EY? oe
—— - x7
n
=1.708
Ox Oy
N= ‘TWordimensional Random Variables © 415
G Caleulate the correlation coeticien
and the lines of
te the following data: TAU December *04, June 06)
ee inal)
e s is 5 ia [1320] 74
¢ a6 125 208
ape
i 339 312
4 40 240
es -26 130
a 20 20
0 0 0
1 -13 -13
2 15 30
a B 172
xy=-13 | ZY =-80 | SAY = 89)
1 -13
45x =2x(-13)=—
t3x 3% 3) a
1
Egy 2 x¢-90) =~ 19
Ley i.
1
Foxy = 5x69) = 111.375
2 alsy2 at ay oe
Oy = LEX! XP =e x (147) — # 7 I:
t5y2
n
; (6440) — (-10)? = 705
-13
1 -| ==) e109)
ged? -A¥ 111.375 (3 Je )
eee p09
Oxo, ~*~ K2GSS id
hhobain the repression lines:
"egression line of x on y is
Pry =
x-¥=p ZO-D
Which is used to Predict or estimate bs value of x for any given y.
-13 ™~
416 © Probability and Random Processes
= -13
x=X SOS aro +70 = 68.375
Y = ¥ +165=-10+165=155
3.96
> = $8.375 = 0.9x——(y- 155
7 2655019)
x — 68.375 = 0.134(y — 155)
x = 0,134y — 20.8 + 68.375
> x= 0.134y + 47.5
The regression line of y on x is
gy _
y-y =P -X)
Oy
26.55
‘y — 155 = 09x 68.375
a : * 3:96 :
y = 6.034x — 412.58 + 155
= y= 6.034x — 257.58
EXAMPLE 5.73 Calculate the correlation coefficient between the variables
x and y from the following data:
x 1 2 3 4 5 6 7
9 8 10 12} 11 13 14 EXAMPLE sy If ¥=970, F=18, 0, = 38,0, =2 and p = 0.6, find the
line of regression of x on y. [AU December ’09]
Solution Given: *¥ = 970, ¥ = 18, 0, = 38, o, = 2, p= 0.6
The regression line of x on y is
= — oO; =
x-¥ = by-H=p—(y-¥)
ey
(& - 970) = o6(38)0v —18)
x= 970 + (0.6) (19) (y— 18)
= 764.8 + 11.4y
+. The regression line of x on y is
x= ll4y + 764.8 TXAMPLE 440 The two regression lines are 4x — Sy + 33 = 0 and
tie 9) = 107, and Var(X) = 25, Find
() the means of x and y
2 the values of p and o,. and
It) the angle between the regression lines.
[AU June ’06, December ’07] 424 & Probability and Random Processes
Solution
i) Given; 4x - Sy =~ 33
ie 20x - Sy = 107
Since all the regression lines pass through the means
4x -—5y =-33
20% —9F = 107
Multiplying Eq. (i) by 5,
20% — 25F =-165
Equation (iii)-Eq. (ii) gives
Substituting in Eq. (i), we get
4x =-33 + 5(17)
85-33 _
4
x= 13
(ii) = 20x - 9y = 107
er oe
> 20
x= O.45y + 5.35
= 0.45 = coefficient of y
oy
Sy=33 + 4x
a .8 = coefficient of x
P=, Xb, = 0.36
The correlation coefficient p = 0.6
iii) Given: g? = 25
r o,=5
b= en o45 0.6x5
=*=0. o, =
%, G8 0.45
= 6.67
|
@ Two-dimensional Random Vari
bles 425
o\ 4 angle between the regression lines are
itiprne
a 9
P= tan'0.512) = ge ay u
1 IFEX, Y denote the deviation Of variance ff
de iW on tom the arithmeti
wat ifp= 05, EXY = 120, 0, = 8, 5x2 = 9p pre "number of times
ot von X= x-¥,Yay-F
win Given
‘ 226-D0-7) Sear
= * :
: t v2 Yop
1 (20)
adler
x9.48 x
vn
120
Saag
Wi Saaxexos
n=10
BUAMPLE 5.82 Can Y = 5 + 2.8¥ and x
= 3 ~ 0.5Y be the estimated
regession equations of ¥ on X and X' on ¥ reg
Pectively? Explain your answer,
[AU June *06, November °07]
Siition Given: X=3-O5Y = by = 95
Y=S428V > 5, =28
P= byy X by = (0.5) x (28) = 1.4
P= V-1.4 which is imaginary quantity.
Pcannot be imaginary,
® The given lings are not estimated as regression equations
EXAMPLE, rat Iy= 2x ~3 and y= Sx +7 are the two regression lines,
fal te mean values of x an
id y. Find the correlation coefficient between
dy Find an estimate of x when y = 1
Souion The tw
© regression lines always pass through their means (¥, 7)
* We have
y= 25-3 426 @ Probability and Random Processes
T4+7 => 2x
Solving we get -10
+. The correlation coefficient Pyy = Pry Pyx =
Aliter
heed
If we choose the regression equation for x as x = >¥ +5 and for yas y=
Sx +7, then by =
s Itis nt a
EXAMPLE 544 Given that x = 4y +5 and y = Ax +4 are the lines of
regression of x on y and yon x respectively, show that 0 < 4k < 1. If k= 1/16,
find the means of two variables and coefficient of correlation between them
Solution Given: x= 4y + 5, regression line of x on y
: by =
‘The regression line of y on x is
yeket4
b.=k
Bu 0s Psl = tl ci aoa
Itk= a then
Pathe tod > path ”
Two-dimensional Random Variables © 427
i positive, is also positive,
yy
abot” a
cs oo
‘ession are
yo ines — Et S
x
=244
2 16
regression lines pass through their means
the
0 x = 4y
¢ z= 47 +5 @
_ x
yard Gi)
jon) + 4% Ea GA gives
Eu
rei+2l = 3F=84 > F228
ating in Eq. (1)
subst 28= 4y+5 = y=5.75
+t}a=3
8
EQ?)
ra 1
xf S 23 ne
I (ade Je sl +4}ac
: alle
= 3/8
~ 215 Two: ‘dimensional Random, Variables © 433
it can be proved that
i
Pe a 2
Var(X) = E(x) — [Eee
= zy _B
1s \g 960
Var() = EW?) ~[eqyp . 23
7
ol
EY) = J I 9 FC y)dxdy
1
=F) foc? +a
00
Cov(x, ¥)
CovKsy) >
Pxr= Ox oy 70
960
1.960 960,
6a 93 = 760 =
Is
Per = 93
“line of regression of X on ¥ is given by
X~ BQ) =
LUX My _ gery
o 434 @ Probability and Random Processes
1
5__/,_5
x Ble 3)
960
-2-5{r-2)
8 73 8
-15 55
XY +
a 23 73
The regression line of Y on Xis
Y
Y-EQ)= Cr Diy — Ec)
ox
EXAMPLE 5.92 Let the random variable X have the marginal densi
Y=L ! < re
f(x)=1, pans
and the conditional density of Y be
1
JO!x) = 1, xeyexth-7ers0
FL -x
