Error "sample may not include multiple panels" for *Durbin Watson test

[Scott Rick]

Hi. I am using panel data and I want to test for autocorrelation in my dataset. My panel identifier is a consultant code, which is unbalanced. My time variable, which is the week number, has gaps. That is, for some weeks I may have no data for some consultants. Each consultant does have at least four panel observations. However, I still get the "sample may not include multiple panels" on running the Durbin Watson test for autocorrelation.

Could someone please help me here?

tsset week_gen consultant_code
       panel variable:  week_gen (unbalanced)
        time variable:  consultant_~e, 1 to 100, but with gaps
                delta:  1 unit



. reg OCbyWeek consultant_code i.week_gen i.sex i.strata_campus i.age_group

      Source |       SS           df       MS      Number of obs   =     1,949
-------------+----------------------------------   F(28, 1920)     =      5.39
       Model |    2771.914        28  98.9969285   Prob > F        =    0.0000
    Residual |  35236.5314     1,920  18.3523601   R-squared       =    0.0729
-------------+----------------------------------   Adj R-squared   =    0.0594
       Total |  38008.4454     1,948  19.5115223   Root MSE        =     4.284

---------------------------------------------------------------------------------
       OCbyWeek |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
----------------+----------------------------------------------------------------
consultant_code |  -.0100498   .0035934    -2.80   0.005    -.0170971   -.0030024
                |
       week_gen |
             2  |   .6503219   .6654128     0.98   0.329    -.6546859     1.95533
             3  |   1.482073   .6634909     2.23   0.026     .1808339    2.783311
             4  |   .3945532   .6652723     0.59   0.553    -.9101791    1.699285
             5  |   .5017673   .6580053     0.76   0.446    -.7887129    1.792248
             6  |   .9866308   .6597859     1.50   0.135    -.3073415    2.280603
             7  |   .6783259   .6562974     1.03   0.301    -.6088048    1.965457
             8  |   .6901914   .6674288     1.03   0.301    -.6187701    1.999153
             9  |  -.0369435   .6580267    -0.06   0.955    -1.327466    1.253579
            10  |   .7780871    .651276     1.19   0.232    -.4991956     2.05537
            11  |     .09302   .6545462     0.14   0.887    -1.190676    1.376716
            12  |   1.005607   .6580206     1.53   0.127     -.284903    2.296117
            13  |   .2673806   .6652555     0.40   0.688    -1.037319     1.57208
            14  |   .5045686   .6545791     0.77   0.441    -.7791923    1.788329
            15  |   .6189594   .6599452     0.94   0.348    -.6753253    1.913244
            16  |   .4646031   .6581353     0.71   0.480    -.8261321    1.755338
            17  |  -.3828458   .6617966    -0.58   0.563    -1.680762      .91507
            18  |   .9784346   .6496361     1.51   0.132    -.2956319    2.252501
            19  |    1.06098   .6496156     1.63   0.103    -.2130461    2.335006
            20  |   1.104494   .6512457     1.70   0.090    -.1727292    2.381717
            21  |   .8141989    .658012     1.24   0.216    -.4762944    2.104692
            22  |  -.8659594    .657982    -1.32   0.188    -2.156394     .424475
                |
            sex |
             M  |  -.1528916   .2051885    -0.75   0.456    -.5553073    .2495241
                |
  strata_campus |
     1  |   1.118929   .2684115     4.17   0.000     .5925202    1.645338
     2  |   2.745076     .38681     7.10   0.000     1.986464    3.503688
 3  |     1.0982   .3883027     2.83   0.005     .3366602    1.859739
                |
      age_group |
             2  |   .9321409   .2336606     3.99   0.000     .4738857    1.390396
             3  |   2.739329   .3225472     8.49   0.000      2.10675    3.371909
                |
          _cons |   .9975297    .569755     1.75   0.080    -.1198739    2.114933
---------------------------------------------------------------------------------

. estat dwatson
sample may not include multiple panels
r(459);

.

[Andrew Musau]

The Durbin-Watson D test is not designed for panel data. See xtserial from the Stata Journal instead.

SJ-3-2  st0039  . . Testing for serial correlation in linear panel-data models
        (help xtserial if installed)  . . . . . . . . . . . . .  D. M. Drukker
        Q2/03   SJ 3(2):168--177
        test for serial correlation in random- or fixed-effects
        one-way models that can be applied under general conditions

findit xtserial

[Jeff Wooldridge]

How many consultants are in the data set? Looks like maybe 100 or so?

[Scott Rick]

Andrew Musau Thanks. I wasn't aware of that

Jeff Wooldridge Yes, there are 100 consultants. Does that have a bearing on anything here?

[Nick Cox]

A important detail here is that the tsset call looks quite the wrong way round. The panel identifier should be mentioned first and the time variable second.

Another important detail is that consultant code is not plausible as a counted or measured predictor in the regression. Whether week may be so considered is more up in the air.