Cong Thuc KTL Tong Hop2
Cong Thuc KTL Tong Hop2
Cong Thuc KTL Tong Hop2
Yi = 1+ 2Xi + Ui
1 =
2 =
- 2
XY X Y
X
( X )
x y
x
i
Se ( 2 ) =
Se ( 1 ) =
X
n x
2
i
+ 2 X0
+> c lng bng khong tin cy:
( E (Y / X = X 0 )) max
= Y0 + se(Y0 )T( n 2)
( E (Y / X = X 0 )) min
= Y0 se(Y0 )T( n 2)
se(Y0 ) =
1 ( X 0 X )2
+
n
n
xi2
se(Y0 ) = 1 +
i =1
i =1
ESS
TSS
R2
1
TSS
n 1
Yi = 1+ 2X2i + + kXki + Ui
n 1
= 1 - (1-R2) n k
1 ( X 0 X )2
+
n
n
xi2
)2
DHKT
0
Y0 = 1 + 2 X 2 +
+ k
X k0
( E (Y / X = X 0 )) max
= Y0 + se(Y0 )T( n k )
( E (Y / X = X 0 )) min
= Y0 se(Y0 )T( n k )
se(Y0 ) =
( X ) cov( ) X
, se(Y0 ) =
( X ) cov( ) X
1
X 0
= 2
...
0
X k
X0
+ Ui
j=
1, k
+ UL im: j ; j =
1, k
< j + Se ( j
)T /2(n-k)
+ ( j)max = j + Se ( j ) T (n-k)
T (n-k)
+ ( j)min = j - Se ( j )
2. Kim nh so snh
DHKT
vi s thc
2
cho trc , j =
1, k
Cp gi thuyt
MBB ca H0
H0 :
H1 :
>
H0 :
H1 :
<
H0 :
H1 :
(n-k)
= { T = j j ; T> T
Se( j )
W = { T=
jj ; T < - T
Se( j )
(n-k)
= {T = j j ; /T/ > T
Se( j )
/2
(n-k)
H0: a
W = {T =
a j b j' c
;T>
Se ( a j b j ' )
T (n-k)}
H1 : a
>c
H0 : a
W = {T =
a j b j' c
; T<-
Se (a j b j ' )
T (n-k)}
H1 : a
<c
H0 : a
W = {T=
=c
a j b j' c
Se (a j b j ' )
T /2(n-k)}
H1 : a
Se (a j
DHKT
b j ) =
'
; /T/>
+ c lng im:
i =1
n k
KTC 2 pha
(n k ) 2
2(/n2k )
<
max
2
= (n 2 (kn)k)
1
2
< (n 2(kn)k)
1 / 2
min
2
= (n 2(kn)k)
MBB ca H0
H0 : 2
(n k ) 2 ; 2 > 2 ( nk ) }
W = { =
2
0
2
02
H1 : 2 > 02
H0 :
2
(n k ) 2 ; 2 < 2 ( nk ) }
W = { =
1
2
0
2
02
H1 : 2 < 02
2
k)
W = { = (n k ) ; 2 > 2 (/n
hoc 2 <
2
2
0
H 0 : = 02
12(n/ 2k ) }
H1 : 2
02
6. Kim nh ph hp hi qui
H0 : R2 = 0 ( (1) khng ph hp )
=
H0 :
=0
2
H1 : R > 0 ( (1) c ph hp )
MBB ca H0 : w = { F =
H1 :
R 2 /( k 1)
(1 R 2 ) /( n k )
j =2
(UR) : Yi =
(R)
+ 2X2i+ + kXki + Ui
: Yi =
H0 : (UR) (R)
+ 2X2i+
k-m
H0 :
k-m+1
(k-m)
X(k-m)i+ Ui
k-m + 2
2
j
; F > F (k-1;n-k)}
DHKT
= =
=0
H1 : (UR) (R)
H1 :
j =k m +1
MBB ca H0: w = { F =
k (UR ) )
( R(UR ) R( R ) ) / m
2
}
MT S dng hi quy ph dng pht hin
cc khuyt tt trong m hnh
Yi = 1 + 2 X2i ++ kXki + Ui (1)
M hnh ph:
j+1
+ 2X2i + +
j-1
X(j-1)i +
H0
- 1
, Xj
+1
Xk )
H1 : R2(a) > 0 ( (1) c a cng tuyn )
Khuyt tt PSSST
UL (1) OLS
ei, Yi
K ParK:
ei2 = e 1
M hnh ph:
Ln e2i =
2
X
ji
e vi
+ 2ln Xji + Vi
H 0 : R2(b) = 0
H0 :
(b)
= 0
(1)
(1)
c PSSS thay i)
i)
H 1 : R2(b) > 0
H1 :
K Glejser:
Cc m hnh ph:
+ / ei/ =
e2i =
+ /ej / =
DHKT
+ 2Xji + Vi
+
+
(c)
Xji2 + Vi
1
X ji
+ Vi
5
+ /ej / =
ji
+ Vi
+ / ej / =
ji
+ Vi
H0 :
H0 : R2(c) = 0
=0
(1)
i)
H 1 : R2(c) > 0
H1 :
(1)
c PSSS thay i)
K White:
M hnh ph:
e2i =
H0 : R2(*) = 0
H1 : R2(*) > 0
K theo dng D
M hnh ph:
H0 :
(*)
(bng MFit):
e2i =
R2(*) = 0
2
i
H0 :
+ Vi
(*)
= 0
(1)
c PSSS
khng thay i )
+>
H1 :
H1 :
R 2(*) > 0
( (1) c
PSSS thay i )
MBB ca H0:
W = { 2 = n(*)
R(*2)
2 (1)
; 2 > }
w = { F =
n(1) -2)
R(*) / 1
2
(1 R(*) ) /( n (*) 2)
; F > F (1;
qs
DHKT
(e
t =2
K Durbin - Watson : d =
et 1 ) 2
e
t =1
2 (1- )
c Kl
>0
0
Khn
= 0
dL
<0
g c
Kl
4-d U
dU
4 - dL
X2i+ + kXki +
k+1
Yi-1+Ui
(1)
n
h=
1 n var( k +1 )
(1-
<0
d
2
1 n var( k +1 )
= 0
> 0
-1,96
1,96
et, et-1
OL S
M hnh ph:
et =
X2t ++ kXkt + Vt
(*)
OL S R(2*)
et =
H 0 : = 0 ( (1) AR (1) )
(**)
OL S R(2**)
H0 : (*) (**)
H1 : 0
2 (1) }
( (1) AR (1) )
W = { 2 = (n(1) 1 )
MBB ca H0:
DHKT
H1 : (*)
2
R(*)
(**)
; 2 >
w = { F =
n(*)
k (*)
( R(*) R(**) ) / 1
2
; F > F (1;
qs
Khuyt tt v dng hm (St bin gii thch)K theo dng B (bng MFit):
Y i = 1+ 2X2i+ + kXki +Ui
(1)
M hnh ph:
+
k+1
Yi + U i
(2)
( R 2 ( 2) R 2 (1) ) / 1
={F=
;F>
2
(1 R( 2) ) /(n(1) k ( 2) )
MBB ca H0: W
F (1 ,
(1)
-k
( 2)
K Lagrange:
ei =
M hnh ph:
2
Yi +Ui
+ 2X2i+ +
(1)
W = { 2 = n(1) . R2(3)
; 2
qs
k+1
(3)
MBB ca H0:
> 2
(1)
(1)
ei, ei2, e3 i
+>
UL
+>
H0 : Ui N (0, 2 )
H1: Ui ~ N (0, 2 )
OL S
S,K
MBB ca H0 : W = { 2 = n [
Hoc s dng P-Value ca
qs
S2
6
( k 3) 2
24
2( 2)
]; 2 > }
ti dng C ca bng bo co
2.4285
?
Adjusted Rsquared
S.E. of regression
.75091
?
?
6.0715
Mean dependent
var
S.D. dependent var
Akaike info
criterion
Schwarz criterion
0.0000
112.8706
13.4243
3.335965
Sum squared
628.4441
3.396482
resid
Log likelihood
-54.8072
F-statistic
25.1168
Durbin-Watson
1.3209
Prob(F-statistic)
0.0000
stat
Serial Correlation LM Test : Kim nh v tng quan
chui:
Breusch-Godfrey Serial Correlation
LM Test:
F-statistic
Probability 0.22799
Obs*R1.863
Probability 0.17200
squared
8
DHKT
10
6.466
Probability
1
Hy cho bit Fqs=6.4661 oc tnh nh th no? (nu cc bc
tin hnh v cng thc tnh). Sau vit cp gi thuyt tong ng ri thc hin c th kim nh rt ra kt lun.
Bi 3:
Cho kt qu hi quy sau:
Dependent Variable: LQ
Method: Least Squares
Sample: 1 15
Included observations: 15
Variable
Coefficient
Std.
t-Statistic
Prob.
Error
C
.84010
.27177
3.0912
0.0000
LK
1.1816
.30204
3.9121
0.0002
LL
.67310
.15314
4.3952
0.0000
R-squared
Mean dependent
9.9492
var
Adjusted R.97952
S.D. dependent var
.56629
squared
S.E. of regression
.081037
Akaike info
DHKT
11
criterion
Schwarz criterion
Sum squared
.078804
resid
Log likelihood
-54.8072
F-statistic
335.8295
Durbin-Watson
1.9981
Prob(F-statistic)
0.0000
stat
Serial Correlation LM Test : Kim nh v tng quan
chui:
Breusch-Godfrey Serial Correlation
LM Test:
F-statistic
.
Probability 0.22799
80876
Obs*R1.027
Probability 0.17200
squared
3
trong Q: mc sn lng, K: s vn u t, L: s lao ng
trong nm ca doanh nghip. L*=Ln(*).Cho =5%.
1- Hy vit m hnh hi quy mu ca cc bin gc (Q,L,K).
2- Gii thch kt qu c lng cc h s hi quy ring. Kt qu
c lng c ph hp vi l thuyt kinh t khng?
3-Vn u t c nh hng n sn lng ca doanh nghip
khng? Nu c th theo chiu hng no?
4-Phi chng khi s lao ng gim 1% th sn lng gim hn
0,8%?
5-Khi tng s vn u t tng 1% th c th hy vng sn lng
doanh nghip tng ti a bao nhiu %?
6-Hy cho bit gi tr DW-statistic = 1.9981 c tnh nh th
no? S dng gi tr kim nh gi thuyt no ca
m hnh? Hy thc hin. C th da vo thng tin no khc
trong bng bo co kt lun v gi thit ny?
Bi tp tng hp
Cho cc bin (n v nghn t ng) GDP: Tng sn phm
trong nc; EX: Tng gi tr xut khu; IM: Tng gi tr nhp
khu; GIP: Tng gi tr sn phm cng nghip; GAP: Tng gi
tr sn phm nng nghip .
Tin hnh hi quy c kt qu m hnh (1) sau:
Cho kt qu hi quy sau:
Dependent Variable: GDP
Method: Least Squares
Sample: 1 17
Included observations: 17
DHKT
12
Variable
C
EX
IM
R-squared
Coefficient
6.7209
-1.6109
Std. t-Statistic
Error
.96634 20.7720
6.3330
.79317 -2.0310
Mean dependent
var
S.D. dependent
var
Akaike info
criterion
Schwarz criterion
Prob.
0.0000
0.0002
28.4235
Adjusted R.93988
9.3501
squared
S.E. of
.081037
regression
Sum squared
73.5831
resid
Log likelihood
-36.5762
F-statistic
126.0677
Durbin-Watson
.67180
Prob(F-statistic)
0.0000
stat
Serial Correlation LM Test : Kim nh v tng quan
chui:
Breusch-Godfrey Serial Correlation LM
Test:
F-statistic
7.8534
Probability 0.01500
3
Obs*R6.4022
Probability 0.01090
squared
9
White Heteroskedasticity : Kim nh v phng sai sai
s thay i
(No cross tems)
White Heteroskedasticity Test:
F-statistic
1.5724
Probabilit 0.214818
y
Obs*RProbabilit 0.187007
squared
y
Ramsey RESET : Kim nh v dng hm
Ramsey RESET Test:
F-statistic
9.6547
Probabi 0.002037
lity
Log likelihood
9.6
Probabi 0.000059
ratio
547 lity
Ma trn v hip phng sai gia c lng cc h s ng vi cc
bin:
INPT
EX
IM
INPT
.93381
.35155
-.41074
EX
DHKT
.35155
1.1263
13
-.80851
IM
-.41074
-.80851
.62911
(4)
(5)
R 2= .0070257
(6)
14
(8)
R 2= .91776
15
Variable
Coefficient
C
LEX
LIM
R-squared
3.1256
.27560
.096379
.98180
Adjusted Rsquared
S.E. of
regression
Sum squared
resid
Log likelihood
Durbin-Watson
stat
.97920
.046213
.029899
Std. t-Statistic
Error
.041453 75.4020
.030889 8.9221
.049027
1.9658
Mean dependent
var
S.D. dependent
var
Akaike info
criterion
Schwarz criterion
F-statistic
Prob(F-statistic)
1.1499
Prob.
0.0000
0.0000
0.0690
3.2983
.029899
377.6539
0.0000
.0010885
.9541E-3
-.0013935
LIM
-.0019493
-.0013935
.0024036
a- Vit m hnh hi quy mu vi cc bin gc. Cho bit
ngha c lng cc h s trong m hnh.
b- Vi cu trc m hnh ny, phi chng mc tc ng ca
EX ti GDP l mnh hn mc tc ng ca IM? Nu EX v
IM cng tng 2% th mc thay i ti a ca GDP l
bao nhiu?
8> Xt 2 thi k ca b s liu. To bin D=0 vi thi k
1980-1985, D=1 vi thi k 1986- 1996; DEX=D*EX. Hi
quy m hnh (11) cho kt qu sau:
Dependent Variable: GDP
Method: Least Squares
Sample: 1980 to 1996
Included observations: 17
Variable
Coefficient
Std. t-Statistic
Prob.
Error
C
12.0882
2.3703
5.0998
[.000]
EX
17.4547
4.4157
3.9529
[.002
IM
-.97188
.52123
-1.8646
[.087]
D
10.1480
2.4735
4.1026
[.001
DHKT
16
DEX
R-squared
-12.1563
.98233
Adjusted Rsquared
S.E. of
regression
Sum squared
resid
Log likelihood
Durbin-Watson
stat
.97644
1.4351
4.3663 -2.7841
Mean dependent
var
S.D. dependent
var
Akaike info
criterion
Schwarz criterion
F-statistic
Prob(F-statistic)
[.017
28.4235
9.3501
166.7895
[.000
DHKT
17