Hi Martin and Johannes,
What I meant was that it seems Stata developed xtdpd based on the pre-
existing xtabond2 written by someone else...so they were developed separately,
but I thought they both went to the same end point.
From the code I attached before you can see that I used the exact same
variables in the equation and in the instruments. Based on Roodman (2006,
updated 2008) I included -dgmmiv(rpci06)- instead of -dgmmiv(L.rpci06)- because
the variable is predetermined.
Please correct me if I am mistaken, but I read [XT] xtdpd to mean that it
performs the full system estimation, as does xtabond2.
I created a table to compare b, se, and p for both estimations:
----------------------------------------
Variable | xtabond2 xtdpd
-------------+--------------------------
L.rpci06 | 1.0082279 .95854292
| .00703641 .01715558
| 0.0000 0.0000
dpop | -10586.382 -9771.2316
| 8322.6778 8789.469
| 0.2034 0.2663
emprate | 931.76409 2463.3689
| 299.83173 490.01287
| 0.0019 0.0000
affilf | 21.256932 64.378555
| 26.493406 28.89868
| 0.4224 0.0259
indf | -.58609293 87.438091
| 37.376563 55.416926
| 0.9875 0.1146
stadnewdum | -57.844387 -65.458247
| 61.194835 56.162188
| 0.3445 0.2438
year2 | 48.506442
| 118.55031
| 0.6824
year3 | -218.09176 -345.9017
| 108.64957 41.142174
| 0.0447 0.0000
year4 | -166.90713 -288.23428
| 101.24846 50.274958
| 0.0993 0.0000
year5 | 60.137151 -76.550558
| 97.009136 41.917769
| 0.5353 0.0678
year6 | -651.36857 -759.32052
| 74.128899 56.315531
| 0.0000 0.0000
year7 | -883.72056 -962.77852
| 63.059393 67.084028
| 0.0000 0.0000
year8 | 153.95923 13.493757
| 79.083477 60.330085
| 0.0516 0.8230
year9 | -593.71338 -680.76975
| 73.279566 56.754457
| 0.0000 0.0000
year10 | -34.730374 -151.11239
| 80.943102 55.31604
| 0.6679 0.0063
year11 | -362.90865 -464.29723
| 81.49597 40.793765
| 0.0000 0.0000
year12 | -13.034162 -119.66874
| 98.41181 54.555359
| 0.8946 0.0283
year13 | 58.446669 1.2973734
| 89.004505 45.862199
| 0.5114 0.9774
year14 | 412.37828 350.77885
| 98.300283 46.139216
| 0.0000 0.0000
year15 | -262.04283 -253.9137
| 97.984314 58.200346
| 0.0075 0.0000
year16 | 106.03241 81.865353
| 101.77577 73.303116
| 0.2975 0.2641
year17 | -672.68695 -608.5532
| 107.01528 81.820516
| 0.0000 0.0000
year18 | -656.81837 -606.05313
| 104.84317 80.886586
| 0.0000 0.0000
year19 | -449.81415 -396.59992
| 115.17053 95.370238
| 0.0001 0.0000
year20 | 66.691583 113.09717
| 95.629426 74.918548
| 0.4856 0.1311
year21 | -369.28923 -271.06565
| 135.80695 111.31033
| 0.0065 0.0149
_cons | -1.0757578 639.66347
| 187.76939 267.14751
| 0.9954 0.0166
----------------------------------------
legend: b/se/p
xtdpd clearly dropped an additional year - perhaps xtdpd is set to default to
using 2 lags while xtabond2 uses only 1 by default?
- Nola
From Johannes Geyer <[email protected]>
To [email protected]
Subject Re: st: AW: xtdpd syntax v. xtabond2 syntax
Date Fri, 27 Feb 2009 13:53:26 +0100
Both programs are similar; but they do not share exactly the same syntax.
If you want them to do the same, you need to control the instrument matrix
and the type of model "difference" vs. "system". If you ask members of the
list to help, it is easier if you show us an estimation output. I guess
the number of instruments differs between your models. xtabond2 defaults
to the system estimator, and you have to specify -noleveleq-, futhermore,
you need to include the "l." in your option -gmm()-.
*****************************
webuse abdata, clear
xtabond2 l(0/2).n l(0/1).(w ys) k, gmm(l.n) iv(l(0/1).(w ys) k) noleveleq
xtdpd l(0/2).n l(0/1).(w ys) k, dgmmiv(n) div(l(0/1).(w ys) k)
****************************
Johannes
[email protected] schrieb am 27/02/2009 09:24:02:
>
> <>
>
> Wait a minute: -xtdpd- is an official Stata command, while -xtabond2-
comes
> from ssc, right? -xtabond2- is acknowledged in the entry [xt]xtdpd, but
to
> expect both commands to behave similarly would be asking too much...
>
>
>
> HTH
> Martin
>
>
> -----Ursprüngliche Nachricht-----
> Von: [email protected]
> [mailto:[email protected]] Im Auftrag von Nola Agha
> Gesendet: Freitag, 27. Februar 2009 02:13
> An: [email protected]
> Betreff: st: xtdpd syntax v. xtabond2 syntax
>
> Hi Statalisters -
> I was under the impression that xtdpd was Stata's new version of
xtabond2.
> If that was true then shouldn't these two statements produce the same
> results?
>
>
> xtdpd rpci06 L.rpci06 dpop emprate affilf indf stadnewdum year1-year21,
> dgmmiv
> (rpci06) iv(dpop emprate affilf indf stadnewdum year1-year21) twostep
vce
> (robust)
>
> xtabond2 rpci06 L.rpci06 dpop emprate affilf indf stadnewdum
year1-year21,
> gmm
> (rpci06) iv(dpop emprate affilf indf stadnewdum year1-year21) twostep
robust
>
> Yet, I'm getting very different standard errors and coefficients. Any
> thoughts?
>
> Thanks,
> Nola
>
>
>
>
>
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
__________________________________
Nola Agha
Department of Sport Management
Isenberg School of Management
121 Presidents Drive, Room 226
University of Massachusetts
Amherst, MA 01003
Ph: 413-545-5637
E-Mail: [email protected]
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
