删除变量UADF检验及模型函数估计doc.docx
《删除变量UADF检验及模型函数估计doc.docx》由会员分享,可在线阅读,更多相关《删除变量UADF检验及模型函数估计doc.docx(18页珍藏版)》请在冰豆网上搜索。
删除变量UADF检验及模型函数估计doc
删除变量|U=(M2/Mo),建立模型:
经检验:
所有序列(变量)二阶(差分)单整,故可以得出他们之间存在协整关系。
一、对所构造的函数形式:
Islnmlclngdplnvfruu进行OLS估计得:
DependentVariable:
LNM1
Method:
LeastSquares
Date:
07/17/11Time:
16:
07
Sample:
19932009
Includedobservations:
17
Variable
Coefficient
Std.Error
t-Statistic
Prob.
c
-0.711069
0.122437
-5.807651
0.0001
LNGDP
0.889746
0.034797
25.56932
0.0000
LNV
0.025290
0.012065
2.096222
0.0600
F
-0.004546
0.000838
-5.424575
0.0002
R
0.011661
0.003710
3.142834
0.0094
UU
0.819789
0.070029
11.70648
0.0000
R-squared
0.999608
Meandependentvar
6.379730
AdjustedR-squared
0.999429
S・D.dependentvar
0.800568
S・E・ofregression
0.019128
Akaikeinfocriterion
-4.804743
Sumsquaredresid
0.004025
Schwarzcriterion
-4.510668
Loglikelihood
46.84032
Hannan-Quinncriter.
-4.775512
F-statistic
5603.099
Durbin-Watsonstat
2.559494
Prob(F-statistic)
0.000000
因为变量LnV不显著,
将其删除
重新估计:
IsInm1'
cIngdpfi
DependentVariable:
LNM1
Method:
LeastSquares
Date:
07/17/11Time:
16:
10
Sample:
19932009
Ineludedobservations:
17
Variable
Coefficient
Std.Error
t-Statistic
Prob.
c
-0.871897
0.108072
-8.067773
0.0000
LNGDP
0.949776
0.022389
42.42107
0.0000
F
-0.005148
0.000892
-5.772612
0.0001
R
0.008510
0.003842
2.214954
0.0469
UU
0.767808
0.074177
10.35100
0.0000
R-squared
0.999451
Meandependentvar
6.379730
AdjustedR-squared
0.999268
S・D.dependentvar
0.800568
S・E.ofregression
0.021665
Akaikeinfocriterion
-4.586298
ruu
LNM1=-0.871897137169+0.949776234674*LNGDP-0.00514818980363*F+0.00851005559974*R+0.767808226881*UU……
(1);
口各个变量均通过显著性水平0.05的检验。
二、对回归的残差序列E1进行平稳性检验:
NullHypothesis:
E1hasaunitroot
Exogenous:
Constant
LagLength:
0(AutomaticbasedonSIC,MAXLAG=3)
t-Statistic
Prob.*
AugmentedDickey-Fullerteststatistic
-5.107454
0.0011
Testcriticalvalues:
1%level
-3.920350
5%level
-3.065585
10%level
-2.673459
其中El为当前模型的残弟序列,由输出结果,可见序列E无单位根,LnMl与各序列协整。
短期动态货币需求函数估计:
按照Hendry建模理论(年度数据从滞后2期开始,季度数据从滞后8期开始删除不显著变量,本实例中我们均采用年度数据,故从滞后2期开始),首先建立一个能够代表数据生成过程(DGP)的口回归分布滞后模型(ADL),然后逐步简化,最后得到包含变量间长期稳定关系的简单的模型。
为使模型尽量包含被解释变量的有效信息,初始阶数设为2。
因样本数冇限,先考虑删除变量uu(-2):
IsInm1cInm1(-1)Inm1(-2)Ingdplngdp(-1)lngdp(-2)ff(-1)f(-2)rr(-1)r(-2)uuuu(-1)
DependentVariable:
LNM1
Method:
LeastSquares
Date:
07/17/11Time:
16:
58
Sample(adjusted):
19952009
Ineludedobservations:
15afteradjustments
Variable
Coefficient
Std.Error
t-Statistic
Prob.
c
-9.055962
0.614180
-14.74480
0.0431
LNM1(-1)
-1.004107
0.109155
-9.198899
0.0689
LNM1(-2)
-2.931779
0.302015
-9.707409
0.0654
LNGDP
2.485006
0.241660
10.28307
0.0617
LNGDP(-1)
-0.698886
0.349553
-1.999368
0.2952
LNGDP(-2)
3.001560
0.229923
13.05466
0.0487
F
0.042536
0.004507
9.438230
0.0672
F(-1)
-0.002784
0.003329
-0.836107
0.5567
F(-2)
0.002340
0.000726
3.224337
0.1915
R
-0.072725
0.009422
-7.718481
0.0820
R(-1)
0.007009
0.003713
1.887663
0.3101
R(-2)
-0.103633
0.007958
-13.02260
0.0488
UU
0.302357
0.051051
5.922635
0.1065
UU(-1)
1.877907
0.168009
11.17742
0.0568
R-squared
0.999998
Meandependentvar
6.548951
AdjustedR-squared
0.999977
S・D.dependentvar
0.686755
S・E.ofregression
0.003324
Akaikeinfocriterion
-9.416429
Sumsquaredresid
1.11E-05
Schwarzcriterion
-8.755582
Loglikelihood
84.62322
Hannan-Quinncriter.
-9.423468
F-statistic
45957.56
Durbin-Watsonstat
3.141988
Prob(F-statistic)
0.003651
二、F(・1)最不显著,
删除它,
加入uu(・2),
IsInm1cInm1(-1)Inm1(-2)Ingdplngdp(-1)lngdp(-2)ff(-2)rr(-1)r(-2)uuuu(-1)uu(-2)
DependentVariable:
LNM1
Method:
LeastSquares
Date:
07/1刀11Time:
17:
01
Sample(adjusted):
19952009
Includedobservations:
15afteradjustments
Variable
Coefficient
Std.Error
t-Statistic
Prob.
c
-12.89349
5.568058
-2.315618
0.2595
LNM1(-1)
-1.635948
0.950277
-1.721548
0.3350
LNM1(-2)
-3.878068
1.621899
-2.391066
0.2522
LNGDP
1.688993
0.953514
1.771335
0.3272
LNGDP(-1)
0.365671
1.794032
0.203826
0.8720
LNGDP(-2)
4.209730
1.566896
2.686668
0.2268
F
0.063352
0.032094
1.973983
0.2985
F(-2)
0.002852
0.000474
6.012757
0.1049
R
-0.076977
0.015734
-4.892565
0.1284
R(-1)
0.002984
0.006767
0.440884
0.7356
R(-2)
-0.132886
0.045488
-2.921369
0.2100
UU
0.166956
0.162842
1.025261
0.4921
UU(-1)
2.550522
1.074927
2.372739
0.2539
UU(-2)
0.450076
0.633246
0.710744
0.6066
R-squared
0.999998Meandependentvar
6.548951
AdjustedR-squared
0.999974
S・D.dependentvar
0.686755
S・E・ofregression
0.003532
Akaikeinfocriterion
-9.295243
Sumsquaredresid
1.25E-05
Schwarzcriterion
-8.634396
Loglikelihood
83.71432
Hannan-Quinncriter.
-9.302282
F-statistic
40712.36
Durbin-Watsonstat
2.861110
Prob(F-statistic)
0.003879
结果显示,模型的参数显著性更差。
结合一、二,构建模型:
IsInm1cInm1(-1)Inm1(-2)Ingdplngdp(-1)lngdp(-2)ff(-2)rr(-1)r(-2)uuuu(-1)
DependentVariable:
LNM1
Method:
LeastSquares
Date:
07/17/11Time:
17:
02
Sample(adjusted):
19952009
Includedobservations:
15afteradjustments
Variable
Coefficient
Std.Error
t-Statistic
Prob.
c
-8.962392
0.556615
-16.10161
0.0038
LNM1(-1)
-0.964646
0.090718
-10.63344
0.0087
LNM1(-2)
-2.734078
0.173178
-15.78771
0.0040
LNGDP
2.352540
0.168192
13.98725
0.0051
LNGDP(-1)
-0.894218
0.239658
-3.731229
0.0649
LNGDP(-2)
3.105674
0.178150
17.43290
0.0033
F
0.040739
0.003651
11.15777
0.0079
F(・2)
0.002820
0.000410
6.884451
0.0205
R
-0.066849
0.005784
-11.55756
0.0074
R(-1)
0.006893
0.003420
2.015449
0.1814
R(-2)
-0.101033
0.006751
-14.96478
0.0044
uu
0.278238
0.038822
7.166955
0.0189
UU(-1)
1.793263
0.123585
14.51032
0.0047
删除LNGDP(・1)、
R(-1)
IsInm1cInm1(-1)Inm1(-2)Ingdplngdp(-2)ff(-2)rr(-2)uuuu(-1)
DependentVariable:
LNM1
Method:
LeastSquares
Date:
07/17/11Time:
17:
08
Sample(adjusted):
19952009
Includedobservations:
15afteradjustments
Variable
Coefficient
Std.Error
t-Statistic
Prob.
c
-9.615586
1.092128
-8.804451
0.0009
LNM1(-1)
-1.213228
0.131889
-9.198865
0.0008
LNM1(-2)
-2.680661
0.350357
-7.651234
0.0016
LNGDP
1.754064
0.111670
15.70763
0.0001
LNGDP(-2)
2.985912
0.345016
8.654416
0.0010
F
0.045525
0.007065
6.443940
0.0030
F(-2)
0.003005
0.000837
3.588955
0.0230
R
-0.057862
0.010737
-5.388866
0.0057
R(-2)
-0.103145
0.013522
-7.628184
0.0016
UU
0.224446
0.074535
3.011259
0.0395
UU(-1)
2.016339
0.223473
9.022734
0.0008
R-squared
0.999976
Meandependentvar
6.548951
AdjustedR-squared
0.999915
S・D.dependentvar
0.686755
S・E.ofregression
0.006317
Akaikeinfocriterion
-7.146120
Sumsquaredresid
0.000160
Schwarzcriterion
-6.626883
Loglikelihood
64.59590
Hannan-Quinncriter.
-7.151651
F-statistic
16544.36
Durbin-Watsonstat
3.567015
Prob(F-statistic)
0.000000
虽然都通过了,
服力,应增加样本容量。
对于LnM2:
同理,LnM2与其他变量二阶单整,故可以对它们进行I■办整关系检验。
构造与LnMl的类似函数,得到估计结果:
IsInm2cIngdpInvfruu
DependentVariable:
LNM2
Method:
LeastSquares
Date:
07/17/11Time:
16:
33
Sample:
19932009
Includedobservations:
17
Variable
Coefficient
Std.Error
t-Statistic
Prob.
c
-0.642088
0.033720
-19.04200
0.0000
LNGDP
1.024316
0.009583
106.8850
0.0000
LNV
-0.004459
0.003323
-1.341862
0.2067
F
・1.45E-05
0.000231
-0.062754
0.9511
R
-0.010236
0.001022
-10.01699
0.0000
UU
0.610529
0.019286
31.65631
0.0000
R-squared
0.999974
Meandependentvar
7.337301
AdjustedR-squared
0.999962
S・D.dependentvar
0.855129
S・E.ofregression
0.005268
Akaikeinfocriterion
-7.383774
Sumsquaredresid
0.000305
Schwarzcriterion
-7.089699
Loglikelihood
68.76208
Hannan-Quinnenter.
-7.354543
F-statistic
84316.59
Durbin-Watsonstat
1.805058
Prob(F-statistic)
0.000000
删除变量LNV、F,重新定义:
IsInm2cIngdpruu
DependentVariable:
LNM2
Method:
LeastSquares
Date:
07/17/11Time:
16:
36
Sample:
19932009
Ineludedobservations:
17
Variable
Coefficient
Std.Errort-Statistic
Prob.
c
-0.609940
0.024676-24.71820
0.0000
LNGDP
1.014727
0.004905206.8710
0.0000
R
-0.009558
0.000888-10.76388
0.0000
UU
0.618246
0.01769434.94077
0.0000
R-squared
0.999969
Meandependentvar
7.337301
AdjustedR-squared
0.999962
S・D.dependentvar
0.855129
S・E.ofregression
0.005264
Akaikeinfocriterion
-7.453626
Sumsquaredresid
0.000360
Schwarzcriterion
-7.257576
Loglikelihood
67.35582
Hannan-Quinncriter.
-7.434138
F-statistic
140753.6
Durbin-Watsonstat
1.519792
Prob(F-statistic)
0.000000
所以估计函数形式为:
LNM2=-0.609940064088+1.01472690347*LNGDP-0.00955847425234*R
+0.618246494611*UU……
(2)
接下來,对当前模型的残并序列E2进行ADF检验:
NullHypothesis:
E2hasaunitroot
Exogenous:
None
LagLength:
0(AutomaticbasedonSIC,MAXLAG=3)
t-Statistic
Prob.*
AugmentedDickey-Fullerteststatistic
-2.870820
0.0071
Testcriticalvalues:
1%level
-2.717511
5%level
-1.964418
10%level
-1.605603
冇结果可知:
E2无单位根,即序列平稳。
公式(3)中各序列存在协整关系。
短期动态模型(修正模型):
IsInm2cInm2(-1)lnm2(-2)Ingdplngdp(-1)lngdp(-2)rr(-1)r(-2)uuuu(-1)uu(-2)(变量的确定应该是根据长期的模型确定的吧,所以没把LnV、F这两个不显著的变量加入到ECM中)
DependentVariable:
LNM2
Method:
LeastSquares
Date:
07/17/11Time:
16:
42
Sample(adjusted):
19952009
Ineludedobservations:
15afteradjustments
Variable
Coefficient
Std.Error
t-Statistic
Prob.
c
-0.970701
0.232131
-4.181701
0.0249
LNM2(-1)
0.458461
0.251274
1.824546
0.1656
LNM2(-2)
-1.032037
0.273756
-3.769919
0.0327
LNGDP
0.885618
0.037914
23.35857
0.0002
LNGDP