刚才发的部分句子排列有些乱，重发一遍异方差检验的Stata命令：

（1）

xtgls depvar indepvars, igls panels （heteroskedastic）

estimates store heterotest

xtgls depvar indepvars

local df = e(N_g) - 1

lrtest heterotest . , df(`df')

. xtgls lny lnK lnH lnh, igls
Iteration 1: tolerance = 0

Cross-sectional time-series FGLS regression

Coefficients: generalized least squares
Panels: homoskedastic
Correlation: no autocorrelation

Estimated covariances = 1 Number of obs = 580
Estimated autocorrelations = 0 Number of groups = 29
Estimated coefficients = 4 Time periods = 20
Wald chi2(3) = 41817.64
Log likelihood = 297.2711 Prob > chi2 = 0.0000

------------------------------------------------------------------------------
lny | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lnK | .8342499 .0110449 75.53 0.000 .8126023 .8558975
lnH | .2598012 .011067 23.48 0.000 .2381104 .2814921
lnh | .1848715 .0507405 3.64 0.000 .0854221 .284321
_cons | -.8035356 .1137101 -7.07 0.000 -1.026403 -.5806678
------------------------------------------------------------------------------

.
. estimates store heterotest

.
. xtgls lny lnK lnH lnh

Cross-sectional time-series FGLS regression

Coefficients: generalized least squares
Panels: homoskedastic
Correlation: no autocorrelation

Estimated covariances = 1 Number of obs = 580
Estimated autocorrelations = 0 Number of groups = 29
Estimated coefficients = 4 Time periods = 20
Wald chi2(3) = 41817.64
Log likelihood = 297.2711 Prob > chi2 = 0.0000

------------------------------------------------------------------------------
lny | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lnK | .8342499 .0110449 75.53 0.000 .8126023 .8558975
lnH | .2598012 .011067 23.48 0.000 .2381104 .2814921
lnh | .1848715 .0507405 3.64 0.000 .0854221 .284321
_cons | -.8035356 .1137101 -7.07 0.000 -1.026403 -.5806678
------------------------------------------------------------------------------

. local df = e(N_g) - 1

.
. lrtest heterotest . , df(`df')

Likelihood-ratio test LR chi2(28) = 0.00
(Assumption: . nested in heterotest) Prob > chi2 = 1.0000

[此贴子已经被作者于2007-5-15 6:00:14编辑过]

minixi 是用LR test

xttest3是用wald test

您可以用二者分別檢定,看結果是否有一致

. xtgls lny lnK lnH lnh ,panels(he) corr(psar1)

Cross-sectional time-series FGLS regression

Coefficients: generalized least squares
Panels: heteroskedastic
Correlation: panel-specific AR(1)

Estimated covariances = 29 Number of obs = 580
Estimated autocorrelations = 29 Number of groups = 29
Estimated coefficients = 4 Time periods = 20
Wald chi2(3) = 15191.74
Log likelihood = 754.5655 Prob > chi2 = 0.0000

------------------------------------------------------------------------------
lny | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lnK | .7206795 .0154327 46.70 0.000 .690432 .750927
lnH | .317322 .0191238 16.59 0.000 .2798401 .354804
lnh | .7383365 .0941736 7.84 0.000 .5537596 .9229134
_cons | -1.639296 .1981074 -8.27 0.000 -2.02758 -1.251013
------------------------------------------------------------------------------

. xttest3

Modified Wald test for groupwise heteroskedasticity
in cross-sectional time-series FGLS regression model

H0: sigma(i)^2 = sigma^2 for all i

chi2 (29) = 2602.54
Prob>chi2 = 0.0000

. xtgls lny lnK lnH lnh

Cross-sectional time-series FGLS regression

Coefficients: generalized least squares
Panels: homoskedastic
Correlation: no autocorrelation

Estimated covariances = 1 Number of obs = 580
Estimated autocorrelations = 0 Number of groups = 29
Estimated coefficients = 4 Time periods = 20
Wald chi2(3) = 41817.64
Log likelihood = 297.2711 Prob > chi2 = 0.0000

------------------------------------------------------------------------------
lny | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lnK | .8342499 .0110449 75.53 0.000 .8126023 .8558975
lnH | .2598012 .011067 23.48 0.000 .2381104 .2814921
lnh | .1848715 .0507405 3.64 0.000 .0854221 .284321
_cons | -.8035356 .1137101 -7.07 0.000 -1.026403 -.5806678
------------------------------------------------------------------------------

. xttest3

Modified Wald test for groupwise heteroskedasticity
in cross-sectional time-series FGLS regression model

H0: sigma(i)^2 = sigma^2 for all i

chi2 (29) = 533.27
Prob>chi2 = 0.0000

请比较？

这两个结果是不是说明都存在异方差？

用net install xttest2　　　如果可以安装那就行了

不能安装　　用 findit xttest2  然后会弹出个窗口　　有个 click here to install　就可以安装了

默认路径 c:\ado\plus

我试了下很都回归　ＸＴＴＥＳＴ３　Ｐ值都是零　　那就是说都存在异方差？

:)

请教大家一下，我按照前面几位朋友提到的LR方法来检验面板异方差如下：
. xtgls  Y1 X Prop dummy_1999 ……, igls panels(heteroskedastic)

------------------------------------------------------------------------------
          Y1 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
           X |  -1.130337   1.36e-10        .   0.000    -1.130337   -1.130337
        Prop |   1.721421   1.38e-10        .   0.000     1.721421    1.721421
  dummy_1999 |   1.845203   1.18e-09        .   0.000     1.845203    1.845203
  dummy_2000 |  -7.856028   1.14e-09        .   0.000    -7.856028   -7.856028
  dummy_2001 |  -4.864307   1.18e-09        .   0.000    -4.864307   -4.864307
  dummy_2002 |  (dropped)
  ……

------------------------------------------------------------------------------
. est store hetero

. xtgls  Y1 X Prop  dummy_1999 ……（结果略）

. local df = e(N_g) - 1

. lrtest hetero . , df('df')

结果却显示：

''' found where number expected
r(7);

我发现第一个回归的第7行被dropped掉了，这是否是造成这一结果的原因，为什么会出现这种情况呢？谢谢！

多谢ls各位~~

按楼上几位提供的LR方法，结果提示“does not contain scalar e(ll)”，怎么解决

看不懂啊~~

又涨了见识，谢谢

这几个检验的程序好像要从网上下载

感觉这个挺难的，学习学习

能否有哪位高人给我们讲一下自相关和异方差在stata中具体如何实现，结果如何判断呢？谢谢

为什么我的结果会报heterotest does not contain scalar e(ll)

这么好的主贴和回帖啊

收藏备用！

我也遇到了这个问题，怎么解决呀？

.
. local df = e(N_g) - 1

.
. lrtest heterotest . , df(`df')

Likelihood-ratio test                                  LR chi2(149)=    266.52
(Assumption: . nested in heterotest)                   Prob > chi2 =    0.0000

. xttest3

Modified Wald test for groupwise heteroskedasticity
in cross-sectional time-series FGLS regression model

H0: sigma(i)^2 = sigma^2 for all i

chi2 (150)  =   4218.98
Prob>chi2 =      0.0000

Prob > chi2 =    0.0000的意思是存在异方差，之后就可以用gls修正了吗？但修正后还是原来的模型吗？比如说原来的是固定效应模型，修正后还是吗？

好东西一起分析

.xttest3
Modified Wald test for groupwise heteroskedasticity
in fixed effect regression model
H0: sigma(i)^2 = sigma^2 for all i
chi2 (7) = 42.77
Prob>chi2 = 0.0000       Presence of heteroskedasticity
The null is homoskedasticity (or constant variance). Above we reject the null and conclude
heteroskedasticity. Type help xttest3 for more details.