Panel specification- choosing the best model

[Aneta Blazejowska]

Hi,
I try to model panel data with N=28 and T=15 (tsset number year). I have one binary variable in my explanatory variables. More, I want to add time dependency.
1) What is better- using time dummies (i.year) or time trend? If time trend, is it enough to add only variable year which can be treated as trend variable?
2) I try: a) pooled regression b)xtreg, fe c)xtreg, re d)xtgls. Is it good to add time dummies or trend in xtgls?
3) My data has heteroskedasticity, serial correlation and cross-correlation. Which method is better- xtreg, vce(robust) or xtgls? In xtgls how can I know which options choose corr(ar1) or corr(psar1)?

Thanks!

[Carlo Lazzaro]

Aneta:
I would go -xtgls-.
As far as time trend is concerned, I would investigate whether a squared term makes sense:

c.year##c.year

As far -xtgls- options are concerned, see -help xtgls-.

[Aneta Blazejowska]

Does c.year##c.year make sense in xtreg, fe ; xtreg, re and xtgls? I try use squared term in xtgls and it is ommited.

[Carlo Lazzaro]

Aneta:
in the following toy-example I cannot replicate your problem:

. webuse invest2

. xtset company time
       panel variable:  company (strongly balanced)
        time variable:  time, 1 to 20
                delta:  1 unit

. xtgls invest market stock c.time##c.time, panels(hetero)

Cross-sectional time-series FGLS regression

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

Estimated covariances      =         5          Number of obs     =        100
Estimated autocorrelations =         0          Number of groups  =          5
Estimated coefficients     =         5          Time periods      =         20
                                                Wald chi2(4)      =     806.96
                                                Prob > chi2       =     0.0000

-------------------------------------------------------------------------------
       invest |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
       market |    .089049   .0082384    10.81   0.000      .072902    .1051959
        stock |   .3917802   .0401465     9.76   0.000     .3130946    .4704658
         time |  -.7960935   3.526787    -0.23   0.821    -7.708469    6.116282
              |
c.time#c.time |  -.0641139   .1688154    -0.38   0.704     -.394986    .2667583
              |
        _cons |  -16.40302   16.92601    -0.97   0.332     -49.5774    16.77136
-------------------------------------------------------------------------------

.

Pleae check your data or post and example/excerpt of your dataset via -dataex- along with what you typed and what Stata gave you back. Thanks.