Scribd
Upload
11 views3 pages

PS2 Sol

Uploaded by

ongaribelia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
11 views3 pages

PS2 Sol

Uploaded by

ongaribelia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Exercises 1-2 EM2

Jin-Young Choi

(1) Given an i.d.d. sample with a sample size N , …nd a bias of the following estimator for
V ar(Y ).
N
1 X
1) se2Y (Yi Y )2
N
i=1

N
1 X
s2Y )
E(e = Ef(Yi Y )2 g
N
i=1
2 2
Ef(Yi Y ) g = E[f(Yi y) (Y y )g ]
2 2
= Ef(Yi y) 2(Yi y )(Y y) + (Y y) g

= V ar(Yi ) 2Ef(Yi y )(Y y )g + V ar(Y )

= V ar(Yi ) 2Cov(Yi ; Y ) + V ar(Y )


N
1 X
Cov(Yi ; Y ) = Ef(Yi y )( Yj y )g
N
j=1
PN
j=1 (Yj y)
= Ef(Yi y) g
N
N
X1
1 2
= Ef(Yi y) + (Yi y) (Yj y )g
N
j6=i=1
N 1
1 1 X 1
= V ar(Yi ) + Cov(Yi ; Yj ) = V ar(Yi )
N N N
j6=i=1
N
X
1
s2Y ) =
E(e fV ar(Yi ) 2Cov(Yi ; Y ) + V ar(Y )g
N
i=1
XN
1 2 V ar(Y )
= fV ar(Y ) V ar(Y ) + g
N N N
i=1
N
1 X N 1 N 1
= V ar(Y ) = V ar(Y )
N N N
i=1
N 1 1
Bias = V ar(Y ) V ar(Y ) = V ar(Y ):
N N

1
(2) Let’s consider a simple linear model

Y = 0 + 1X + U;

from the population. Assume E(X) = E(U ) = 0 and E(U jX) = 0.


1) Prove
Cov(X; Y )
= 1:
V ar(X)

Cov(X; Y ) E(XY ) E(X)E(Y ) E(XY )


= = * E(X) = 0
V ar(X) V ar(X) V ar(X)
E(X( 0 + 1 X + U )) E(X) + 1 E(X 2 ) + E(XU )
= = 0
V ar(X) V ar(X)
= 1

* E(XU ) = EfXE(U jX)g = 0 & V ar(X) = E(X 2 ) E(X)2 = E(X 2 )

2) For the same Y and X, we consider another linear model

Y = 0 + 1Z
e;
+U where Z e jZ) = 0.
3X and E(U

Find 1 as a function of 1.

Cov(Z; Y ) E(ZY ) E(Z)E(Y ) E(ZY )


= = * E(Z) = 3E(X) = 0
V ar(Z) V ar(Z) V ar(Z)
EfZ( 0 + 1 Z + U e )g 2 e
0 E(Z) + 1 E(Z ) + E(Z U )
= =
V ar(Z) V ar(Z)
= 1 * E(Z Ue ) = EfZE(U e jZ)g = 0 & E(Z 2 ) = V ar(Z)
E(ZY ) 3E(XY ) 1
= = 1:
V ar(Z) 9V ar(X) 3

2
3) For the same Y and X, we consider another linear model

b; 1 b jX) = 0.
W = 0 + 1X +U where W Y and E(U
3

Find 1 as a function of 1.

Cov(X; W ) E(XW ) E(X)E(W ) E(XW )


= =
V ar(X) V ar(X) V ar(X)
EfX( 0 + 1 X + Ub )g 2 b
0 E(X) + 1 E(X ) + E(X U )
= =
V ar(X) V ar(X)
= 1 * E(X Ub ) = EfXE(U b jX)g = 0 & E(X 2 ) = V ar(X)
E(XW ) 1 E(XY ) 1
= = = 1:
V ar(X) 3 V ar(X) 3

You might also like