那个命令是不能做面板数据的

而且才3年的数据

Title

    [R] regress postestimation time series -- Postestimation tools for regress with time series

Description

    The following postestimation commands for time series are available after regress:

    command              description
    --------------------------------------------------------------------------------------------------------
    estat archlm         test for ARCH effects in the residuals
    estat bgodfrey       Breusch-Godfrey test for higher-order serial correlation
    estat durbinalt      Durbin's alternative test for serial correlation
    estat dwatson        Durbin-Watson d statistic to test for first-order serial correlation
    --------------------------------------------------------------------------------------------------------

Special-interest postestimation commands

    These commands provide regression diagnostic tools specific to time series.  You must tsset your data before
    using these commands.

    estat archlm tests for time-dependent volatility.  estat dwatson, estat durbinalt, and estat bgodfrey test for
    serial correlation in the residuals of a linear regression.  For non-time-series regression diagnostic tools,
    see [R] regress postestimation.

    estat archlm performs Engle's LM test for the presence of autoregressive conditional heteroskedasticity.

    estat bgodfrey performs the Breusch-Godfrey test for higher-order serial correlation in the disturbance.  This
    test does not require that all the regressors be strictly exogenous.

    estat durbinalt performs Durbin's alternative test for serial correlation in the disturbance.  This test does
    not require that all the regressors be strictly exogenous.

    estat dwatson computes the Durbin-Watson d statistic to test for first-order serial correlation in the
    disturbance when all the regressors are strictly exogenous.

帮助的上面也写了是regress postestimation 的检验，而不是面板panel的检验。

--------------------------------------------------------------------------------------------------------------------
help for xtserial                                                                                    (SJ3-2: st0039)
--------------------------------------------------------------------------------------------------------------------

Wooldridge test for serial correlation in panel-data models

        xtserial depvar [varlist] [if exp] [in range] [, output]

    You must tsset your data before using xtserial; see help tsset.

Description

    xtserial implements a test for serial correlation in the idiosyncratic errors of a linear panel-data model
    discussed by Wooldridge (2002).  Drukker (2003) presents simulation evidence that this test has good size and
    power properties in reasonable sample sizes.

    Under the null of no serial the residuals from the regression of the first-differenced variables should have
    an autocorrelation of -.5.  This implies that the coefficient on the lagged residuals in a regression of the
    lagged residuals on the current residuals should be -.5.  xtserial performs a Wald test of this hypothesis.
    See Drukker (2003) and Wooldridge (2002) for further details.

Options

    output specifies that the output from the first-differenced regression should be displayed.  By default, the
        first-differenced regression output is not displayed.

Examples

    . xtserial ln_wage age* ttl_exp tenure* south, output

References

    Drukker, D. M. 2003.  Testing for serial correlation in linear panel-data models.  Stata Journal (3)2:
        168-177.

    Wooldridge, J. M.  2002.  Econometric Analysis of Cross Section and Panel Data.  Cambridge, MA: MIT Press.

谢谢蓝版：）

不过我不是面板数据，我的样本是混合截面数据（三年的数据混合）。

我regress y x1 x2后，输入“estat dwatson”或“estat durbinalt”，

结果窗口显示“time variable not set, use -tsset varname ...-”。

然后我就用tsset设置tsset id year。（估计是这设置有问题？）

要怎么处理呢？

谢谢啦。。。

你的数据保证每年对于一个样本就没有问题的。

如果你是每年对于不同的样本，那也会出现问题的。