BASIC ECONOMETRICS

- Econo + Metric

phân tích,
-

2.1.
-
chúng.
2.2.
-
-

+
2.3.
-
-

2.4.
-

-
-
-
2.5.
-

3.1.
-
-
-
3.2.
 -
- Mua
-
3.3.
-
-
1. Regression)

- dependent variable) : Y
- regressor(s)) : X, X2, X3
-

X = Xi (Y/Xi)

- (Estimate
- (Hypothesis testing
- (Forecast, Prediction
(*)

- x !y
- xy
- Y

-
- Y X
(PRF : Population Regression Function).
X = Xi (Y/Xi)
F(Y/Xi)
! E(Y/Xi)
Xi ! E(Y/Xi)
E(Y/Xi) = f(Xi) E(Y/X) =
f(X)

:
E(Y/X) = 1 + 2X
Thì 1 = E(Y/X = 0): (INPT : intercept term)
 E (Y / X )
2 = : (slope coefficient)
X

- Yi (Y/Xi Yi E(Y/Xi)
- ui = Y i E(Y/Xi) : là random
errors)
-
E(ui) = 0 i

- j thì không
.
-
- W = {(Xi, Yi), i = 1÷ n n, n quan sát (observation).
(SRF : Sample Regression Function)
W
, Y = f (X ) (SRF).

E(Y/Xi) = 1 + 2Xi

Yi = 1 + 2 Xi
- 1 và 2 j

nhiên.
- w n, j

Yi Yi ei = Yi Yi (residual).
ei ui
Yi , 1 , 2 , ei E(Y/Xi), 1, 2, u i.
E(Y/Xi)= 1+ 2 Xi
Yi = 1+ 2 Xi + ui

Yi = 1 + 2 Xi
Yi = 1 + 2 Xi + ei

E(Yi) = 1 + 2 X2i + 3X3i kXki

Yi = 1 + 2 X2i + 3X3i kXki + ui
Yi = 1 + 2 X2i + 3 X3i k Xki

Yi = 1 + 2 X2i + 3 X3i k Xki + ei

1. Mô hình
- E(Y/Xi)= 1+ Xi2
Yi = 1+ 2 Xi + ui

- n : W = {(Xi, Yi), i = 1÷ n}, tìm 1 , 2 sao cho SRF: Yi = 1 + 2

Xi

2. Ordinary Least Square)

2.1.
n n
2
- Tìm 1, 2 sao cho (Yi Yi ) 2 ei min
i 1 i 1

XY XY
2 = 2 1 =Y 2 X
X ( X )2
n
xi y i
i 1
xi = Xi X; yi = Yi Y 2 n
xi2 y i
i 1

1 , 2

1 và 2.

- :
n n n
Y Y ei 0 Yi ei 0 ei X i 0
i 1 i 1 i 1

2.2.

E(ui) = 0 i
2
Var(ui) = Var(uj) =
j
Cov(ui, uj) = 0
Cov(ui, Xi) = 0 i

Gauss-Markov: ên t

2.3.
j

E( 1 )= 1 E( 2 )= 2
 n
X i2 1
i 1 2 2
Var ( 1 ) n
; Var ( 2 ) n
n xi2 xi2
i 1 i 1

Se( j ) = Var ( j ) (j = 1,2)

: Cov( 1 , 2 ) XVar ( 2 ).
n
ei2
2 2 2
: = i 1
n k
k
2
= là : (Se. of Regression)

2
ui N(0; )
2
(n k ) 2 2
j N( j ; Var( j )); 2
(n k ) ; Yi N( 1 2 Xi , ).

-
i.
j Se( j )t /2(n k) < j < j + Se( j )t /2(n k)
j < j + Se( j )t (n k)

j Se( j )t (n k) < j

2 2
(n k ) 2 (n k )
KTC 2 phía: 2
< < 2
/ 2 (n k) 1 / 2 (n k)
2
2 (n k )
2
1 (n k )
2
(n k ) 2
2
(n k )

0
*
H0 : j j
* Tqs > t /2(n k)
H1 : j j

* *
H0 : j j j j
* Tqs = Tqs > t (n k)
H1 : j j Se( j )
 *
H0 : j j
* Tqs < t (n k)
H1 : j j

H0 : 2 0 2
Tqs =
H1 : 2 0 Se( j )
* Dùng P value
*
H1 : j j P value P(t tqs )
*
H1 : j j P value P(t tqs )
*
H1 : j j P value 2P(t | tqs |)
P value H0
P value H0 .

0
2 2
H0 : 0
2 2
H1 : 2 2 qs 1 /2 (n k )
0
2 2
qs /2 (n k )
2 2 2
H0 : 0 2 (n k )
2 2
2 2 qs 2 (n k )
H1 : 0 0
qs

2 2
H0 : 0
2 2
H1 : 2 2 qs 1 (n k )
0

Chú ý
H0
+) Chú ý khi j âm.

R2
yi Yi Y n n n
yi Yi Y yi = y i + ei ; y 2i y 2i e 2i
i 1 i 1 i 1
ei Yi Y
TSS = ESS + RSS
TSS (Total Sum of Squares
ESS (Explained Sum of Squares
).
RSS (Residual SS

ESS RSS
R2 = 1 ,0 R2 1
TSS TSS
 2
:

H0 : R 2 0
H1 : R 2 0
ESS /( k 1) R2 n k
Fqs =
RSS /( n k ) 1 R2 k 1
- Fqs > F (k - 1; n - k 0
-
Chú ý: k 2
2
H0 : R 0 H0 : 2 0
-
H1 : R 2 0 H1 : 2 0

- : Fqs = ( 2
)2 .
Se( j )

X = X0
a.
Y0 Se( Y0 )t /2(n k) < E(Y/X0) < Y0 + Se( Y0 )t /2(n k)
1 (X 0 X )2
Y0 = 1 + 2 X0 và Se( Y0 ) =
n xi2
b.
Y0 Se(Y0)t /2(n k) < Y0 < Y0 + Se(Y0) t /2(n k)
1 (X0 X )2
Se(Y0) = 1
n xi2
1. Mô hình
Y k X2, .. ,Xk có

E(Yi) = 1 + 2 X2i + 3X3i kXki (1)

Yi = 1 + 2 X2i + 3X3i kXki + ui (2)
W = {(X2i, X3i ki, Yi); i = 1 n

Yi = 1 + 2 X2i + 3 X3i k Xki (3)

Yi = 1 + 2 X2i + 3 X3i k Xki + ei (4)

*
Y1 = 1 + 2 X21 kXk1 + Y1 1 X 21 ... X k1 u1
1
u1 Y2 1 X 22 ... X k2 u2
2
Y2 = 1 + 2 X22 kXk2 + ... ... ... ... ... ...
...
u2 Yn 1 1 X 2 n 1 ... X kn 1 un 1
k
Yn 1 X 2 n ... X kn un
Yn-1= 1 + 2 X2n-1 kXkn-1 + Y(n = X(n +
1) k) (k 1)
un-1
U(n 1)
Yn = 1 + 2 X2n kXkn +
un

Y=X +U E(Y) = X
Y1 e1
1
Y2 e2
Y= ... ; = 2
;e= ... , thì Y=X
...
Yn 1 en 1 Y=X +e
k
Yn en

2
2
n
Tìm sao cho ei2 = min (Y - X Y-X ) min =
i 1

)-1 thì =( )-1

2
Gt1 : X
Gt2 : E(U) = 0
Gt3 : Var(ui) = 2 i
2
Gt4 : Cov(ui, uj) = 0 Cov(U) = I (I:
Gt5 : Cov(ui, Xi) = 0 i
 Gt6 r(X) = k
=( )-1
2.3. Các tham
E( ) =

Var ( 1 ) Cov( 1 , 2 ) ... Cov( 1 , k )

Cov( 2 , 1 ) Var ( 2 ) ... Cov( 2 , k )
Cov( ) = = 2
( )-1
... ... ... ...
Cov( k , 1 ) Cov( k , 2 ) ... Var ( k )
2 2 e'e
=
n k

3
3
i.
j Se( j )t /2(n k) < j < j + Se( j )t /2(n k)
j < j + Se( j )t (n k)

j Se( j )t (n k) < j

ii.
( i j ) Se( i j )t /2(n k) < i j <( i j ) + Se( i j )t /2(n k)
Se( i j ) = Var ( i j )= Var ( i ) 2Cov( i , j ) Var ( j )
iii.
2 2
(n k ) 2 (n k )
KTC 2 phía: 2
< < 2
/ 2 (n k) 1 / 2 (n k)
2
2 (n k )
2
1 (n k )
2
(n k ) 2
2
(n k )

0
*
H0 : j j
* Tqs > t /2(n k)
H1 : j j

* *
H0 : j j j j
* Tqs = Tqs > t (n k)
H1 : j j Se( j )
 *
H0 : j j
* Tqs < t (n k)
H1 : j j

H0 : i j a i j a
Tqs = Tqs > t /2(n k)
H1 : i j a Se( i j)

0
2 2
H0 : 0
2 2
H1 : 2 2 qs 1 /2 (n k )
0
2 2
qs /2 (n k )
2 2 2
H0 : 0 2 (n k )
2 2
2 2 qs 2 (n k )
H1 : 0 0
qs

2 2
H0 : 0
2 2
H1 : 2 2 qs 1 (n k )
0

4
4
ESS
R2 =
TSS

n 1
R2 = 1 (1 R2) R2 < R2
n k
4
H0 : R 2 0 H0 : 2 ... k 0
H1 : R 2 0 H1 : j 0 : ( j 1)
ESS /( k 1) R2 n k
Fqs =
RSS /( n k ) 1 R2 k 1

Fqs > F (k - 1; n - k 0

4
m Xk-m+1 Xk Y
H0 : k m 1 k m 2 ... 0 k

H1 : j 0: ( j k m 1 k)

E(Y/X2,..,Xk - m,..,Xk ) = 1 + 2X2 kXk (L)

 E(Y/X2 k - m) = 1 + 2X2 kXk - m (N)
2 2
RSS N RSS L n k R R
L N n k
Fqs =
RSS L m 1 RL2 m
Fqs > F (m, n k 0

- m = 1: Fqs = (Tqs)2 Tqs

- m = k 1 : Fqs Fqs

5
i.
Y0 Se( Y0 )t /2(n k) < E(Y/X0) < Y0 + Se( Y0 )t /2(n k)
Y0 = X0 và Se( Y0 ) = X0 ' (X' X) 1 X0
ii.
Y0 Se(Y0)t /2(n k) < Y0 < Y0 + Se(Y0) t /2(n k)
Se(Y0) = 1 X0 ' (X' X) 1 X0

6
6 chi tiêu
6
6.3. Hàm chi phí
6
Y = 0X2 2 X3 3
lnY = ln 0 + 2lnX2 + 3lnX 3
Xét mô hình LY = 1 + 2 LX2 + 3LX3 + v
E(Y / X2 , X3) = e 1X2 2X3 3
1
1 : E(Y/X2 = X3 = 1) = e

2 = E(Y)/X2 : Khi X2 E(Y 2

%
E(Q) = e 1K 2L 3
6.5. Hàm chi phí
6.6. Hàm phân tích
- (qualitative

Y
1
D=
0
Mô hình : E(Y/D) = 1 + 2D
E(Y/D = 1) = 1 + 2
E(Y/D = 0) = 1

2
D (dummy variable).

-
-
-
m

Y X
E(Y) = hsc + hsg.X
1 và A 2.
1 quan sát A1
D=
0 quan sát A1

E(Y/X, D) = 1 + 2X + 3D
 E(Y/X, D) = 1 + 2X + 3DX

E(Y/X, D) = 1 + 2X + 3D + 4DX
H0 : 3 4 0
2 2
H1 : 3 4 0 thái

E(Y) = 1 + 2X
Trong A1 : E(Y) = 1 2
Trong A2 : E(Y) = 1 2

H0 : [ 1 1 1] và [ 2 2 2
H1 : ]
[ 1 1 2 2 ]

W1 n1 trong A1 RSS1
W2 n2 trong A2 RSS2
W = W1 W2 kích th n1 + n 2 RSS
RSS = RSS1 + RSS2.
RSS RSS n1 n2 2k
Fqs = qs > F (k ; n1 + n2 2k) : bác
RSS k
H0
Fqs này và Fqs

X = X*
1 :X X*
D=
0 :X X*
E(Y/X, D) = 1 + 2X + 3( X X*)D
1.
Xét mô hình: E(Yi) = 1 + 2 X2i + 3X3i kXki
GT6:
).
(Multicollinerity).
a. : j j
1 + 2 X2i kXki = 0 i

b. : j 1) sao cho:
1 + 2 X2i kXki + vi = 0 ,
vi
2. Nguyên nhân

-
-
-
-

-
-
-
-
4
4

4
Xj dùng
mô hình (auxilliary regression)
Xj = 1 + 2X2 j-1Xj -1 + j+1Xj+1 (*)
2
H0 : R * 0
2
H1 : R * 0
R*2 n k*
Fqs = ; Fqs > F (k* 1, n k* 0.
1 R*2 k* 1

(* )

4.3.

Xj R2 j
k
m = R2 (R 2 R 2 j )
j 2

5
-
-
- hình
 Yi = 1 + 2 Xi + ui
2
Gt 3 Var(ui)
-
homoscocedasticity).
2
Var(ui) = i
(heterscocedasticity).
2. Nguyên nhân
-
-
-

-
-

4
Var(ui) = i
2
ei2
4
ei, ei ei2
4
2
Gt : i = 2Xi ,
ei2 = 1 + 2Xi + vi (*)
H0 : 2 0 : R*2 0
H1 : 2 0 : R*2 0

2 2
Gt : i = Xi2 ei2 = 1 + 2Xi
2
+ vi
2 2 Xi Xi
Gt : i = ei2 = 1 + 2 + vi
1 1
2 2
Gt : i = Xi ei2 = 1 + 2 X i + vi

C ei Se(ui

4
2 2 2
i = Xi lnei2 = 1 + 2lnX i + vi
 4

Yi = 1 + X2i + 3X3i + ui
2
e2 = 1 + 2X2 + 3X3 + 4X2
2
+ 5X3
2
+ 6X2X3

i (*)
H 0 : R*2 0
H1 : R*2 0
2 2
: qs nR*2 2
qs
2
(k* 1) 0

2 2
i = E(Yi)2
B1 ei Yi
2
B2 ei = 1 + 2 Yi
2 + vi (*)
H0 : 2 0 H 0 : R*2 0 2
H1 : 2 0 H1 : R*2 0
2 2
: qs nR*2 2
qs
2
(1) 0

5.

2
5 i

i
Yi 1 Xi ui
1 2 Yi 1X0i + 2Xi + ui
i i i i

Var(ui
2
5 i

Gt : 2
= 2
Xi Xi
i

Yi 1 ui 2
1 2 Xi
Xi Xi Xi
2 2
Gt : i = Xi2 Xi
2 2
Gt : i = E(Yi)2 Yi
1.

Mô hình Yt = 1 + 2 Xt + ut
Gt 4
Cov(ui, uj) = 0 ( Cov(ut , ut - p) = 0 ( 0)
p (Autocorrelation order p)
p=1
ut và ut-1
ut = ut - 1 + t ( - 1 1, t
=-1
-1< <0
=0
0< <1
=1
T p : ut = 1 ut - 1 + 2 ut - 2 + + p ut - p + t p

2. Nguyên nhân
-
-
-
-
3.

4.
4. Watson
ut = ut -1 + t
n n
2
(et et 1 ) et et 1
t 2 i 1
d= n
2( 1 - n
2
e
t et2
t 1 i 1

-1 1 0 d 4
n, k k dL và dU

Không có Không có
âm
>0 =0 <0
0 dL dU 2 4 dU 4 dL
 4

Chú ý.

Durbin-Watson h:
: Yt = 1 + 2Xt + 1 Yt -1 + ut hay Y= 1 + 2X + 1 Y(-1) + u
n d
h= ; =1-
1 nVar( 1 ) 2

âm
<0 =0 >0
- 1.96 1.96
4

ut = ut -1 + ut -2 ut-p + t
Mô hình et = ( 0) + 1et -1 pet-p + vt (*)
H 0 : R*2 0 H0 : 1 ... p 0 p
H1 : R*2 0 H1 : j 0 : ( j 0)

4
Mô hình
et = [ 1 + 2Xt ]+ 1et -1 pet-p + vt (*)

H0 : 1 ... p 0
H1 : j 0:( j 0)
2 2
: qs n* R*2 (n p) R*2 2
qs
2
( p) 0

et = [ 1 + 2Xt ] + vt (**)
R*2 R*2* n* k*
Fqs = Fqs > F ( k* 1; n* k* 0
1 R*2 k* 1

5.

.
Yt = 1 + 2 Xt + ut
ut = ut -1 + t 0, t
5.1. Khi
Yt = 1 + 2 Xt + ut
Yt -1 = 1 + 2Xt -1 + ut -1 Yt -1 = 1 + 2 Xt -1 + ut -1
Yt Yt-1 = 1(1 )+ 2(Xt Xt -1) + ut ut-1
quát)
*
Yt* = 1 + 2Xt
*
+ t
*
* 1
1 1 = và 2
1
a. =1
Yt = 2 Xt + t
b. = 1
Yt Yt 1 Xt Xt 1
1 2 t (mô hình trung bình
2 2

5.2. Khi

-Watson
d
d 2( 1 - ) =1-
2

et = ( 0) + 1et - 1 + vt 1

-Orcutt
(1) (1)
Yt = 1 + 2 Xt + ut 1 , 2 , et(1)
(1)
et = 0 + 1et-1 + vt 1
(1) ( 2) (2)
1 1 , 2 ,
et( 2)
( 2)
et = 0 + 1et-1 + vt 1

(2)
1

1 và 2 1 và 2 .
1.
-
-
-

3
: Y= 1 + 2X + u (1)
a.
B1 Y
B2
Y=[ 1 + 2X ]+ 1Y
2
mY
m 1 + u (2)
H0 : 1 ... m 0
H1 : j 0, j 1, m
R(22) R(21) n k ( 2)
Fqs =
1 R(22) k ( 2) 1
Fqs > F (k(2) 1; n k(2) 0
b.
B1 e Y
B2
e=[ 1 + 2X ]+ 1Y
2
mY
m 1 + v (*)
H0 : 1 ... m 0
H1 : j 0, j 1, m
2 2
: qs nR*2 2
qs
2
(m) 0.

H0 Normality distribution)
H1
2 2 S2 ( K 3) 2
: JB = qs n
6 24
S skewness), K kutosis)
2 2
qs (2) 0