Stata using xtreg for cluster random effects models

Question

I want to use xtreg to get the random effects intercepts for individual groups and their predicted values. However, all of the predict commands are just populating all of the groups with the constant value.

Using 1 year of data:

xtset group // set panel
xtreg outcome, re
predict u
predict xb

Nick Cox Nick Cox · Accepted Answer · 2017-02-10T18:10:33

This is partly a statistics question. The xt context here makes no difference to a standard feature of regression: if no predictors are specified, then the prediction returned is necessarily constant and equal to the mean response. We can't see your data, but the principle is universal. Here is a reproducible example:

. webuse grunfeld

. xtset
       panel variable:  company (strongly balanced)
        time variable:  year, 1935 to 1954
                delta:  1 year

. xtreg mvalue

Random-effects GLS regression                   Number of obs     =        200
Group variable: company                         Number of groups  =         10

R-sq:                                           Obs per group:
     within  = 0.0000                                         min =         20
     between = 0.0000                                         avg =       20.0
     overall = 0.0000                                         max =         20

                                                Wald chi2(0)      =          .
corr(u_i, X)   = 0 (assumed)                    Prob > chi2       =          .

------------------------------------------------------------------------------
      mvalue |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       _cons |   1081.681   422.1377     2.56   0.010     254.3064    1909.056
-------------+----------------------------------------------------------------
     sigma_u |  1332.6401
     sigma_e |  348.51426
         rho |  .93598465   (fraction of variance due to u_i)
------------------------------------------------------------------------------

. predict predict
(option xb assumed; fitted values)

. su predict

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
     predict |        200    1081.681           0   1081.681   1081.681

. su mvalue

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
      mvalue |        200    1081.681     1314.47      58.12     6241.7

Note that the names you use (examples in your case xb and u) have no implications for what is produced by predict. The default of predict in either case, with no predictors specified, is the predicted mean outcome.

You should check out the options of predict in this case, such as

predict u, u

Stata using xtreg for cluster random effects models

1 Answers