If anybody knows the answer to Kit Baum's question (or possible
explanation), please let me know, I went back to the dpd manual and
couldn't find the answer. If you look at the description of the command,
it simply states: use robust variances in one-step estimator.
g
Is it possible that DPD (even with the basic OLS option) is doing
Newey-West rather than just White (taking autocorrelation into account as
well as heteroskedasticity)? reg,robust in Stata does just plain White
(Huber,sandwich). The newey command, lag(k) would be the equivalent of
Newey-West HAC vce.
Kit Baum
Boston College Economics
--On Tue, Feb 22, 2000 14:31 -0500 Bill Frechete <[log in to unmask]> wrote:
> Once again I have a problem. I ran basic OLS with robust standard errors
> using the dpd package, and stata and I get different results. Is it:
>
> 1. Because they are not implemented the same way.
> 2. I made a mistake and you can tell me what it is.
> 3. None of the above.
>
> I have included the results for both programs to help:
>
> Ox
>
> ------------- One-step estimation using DPD -------------
> (Using robust variance-covariance matrix)
> Dependent variable: corr
> Transformation used: none
>
> Coefficient Std.Error t-value t-prob
> y -0.259338 0.05084 5.10 0.000
> ps -0.00952637 0.009388 1.01 0.311
> pol 0.257687 0.1189 2.17 0.030
> import -0.535737 0.6647 0.806 0.420
> export 1.53499 0.5457 2.81 0.005
> Constant 5.54322 1.114 4.97 0.000
>
> R^2 0.5226714
> no. of observations 994 no. of parameters 6
> RSS 2864.3293661 TSS 6000.749013
> ESE 1.70268 ESE^2 2.899119
>
> constant: yes time dummies: 0
> number of individuals 92
> longest time series 16 [82 - 97]
> shortest time series 1 (unbalanced panel)
>
> Wald (joint): Chi^2(5) = 142.3 [0.000] **
> Wald (dummy): Chi^2(1) = 24.75 [0.000] **
>
> Stata
>
> . reg corr y ps pol import export, robust
>
> Regression with robust standard errors Number of obs =
> 994
> F( 5, 988) =
> 208.44
> Prob > F =
> 0.0000
> R-squared =
> 0.5227
> Root MSE =
> 1.7027
>
> -------------------------------------------------------------------------
> ----- | Robust
> corr | Coef. Std. Err. t P>|t| [95% Conf.
> Interval]
> ---------+---------------------------------------------------------------
> ----- y | -.259338 .0169053 -15.341 0.000 -.2925124
> -.2261635
> ps | -.0095264 .0036233 -2.629 0.009 -.0166367
> -.0024161
> pol | .2576867 .0432606 5.957 0.000 .1727934
> .34258
> import | -.5357368 .396196 -1.352 0.177 -1.313219
> .2417455
> export | 1.534994 .2413327 6.360 0.000 1.061411
> 2.008577
> _cons | 5.54322 .426909 12.985 0.000 4.705467
> 6.380972
> -------------------------------------------------------------------------
> -----
>
>
> Thanks in advance
>
> g
>
> ********** Please do not reply to this address **********
> ******** use [log in to unmask] ********
> ************************ Thank you **********************
>
-----------------------------------------------------------------
Kit Baum [log in to unmask] http://fmwww.bc.edu/ec-v/baum.fac.html
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|