Thanks Mark,
the idea to create the variables manually could indeed solve the
problem. I will try that. Thanks. I will also report the results to
the list, which may take some weeks, however, as the research data
center has its well-earned Christmas holiday now. As a sidenote, the
strange thing is that the command worked with L2 (see the first
version where I used first and second lags), but not when I add third
lags. That makes me think that maybe the problem is not with the lags
but with the combination of estimating the equation in first
differences and the -first- option (or this combined with using 2sls).
However, I would be suprised to learn that I am the first person to
estimate a differenced equation with 2sls and want to have a look at
the first stage, so others should have encountered the same problem
before.
Best,
Nils
On Thu, Dec 17, 2009 at 4:17 PM, Schaffer, Mark E <[email protected]> wrote:
> Nils,
>
> I will have a look at this and see if I can replicate it. But have you tried the very simple fix of creating the L2 variables as new variables instead of having xtivreg2 do them?
>
> That is, instead of
>
> xtivreg2 `depvar' `indepvars' `exovars' (overall_coeff_var
> overall_coeff_var_sq = L2.overall_coeff_var L2.overall_coeff_var_sq
> L3.overall_coeff_var L3.overall_coeff_var_sq), fd cluster(idnum) first
>
> you do
>
> gen double L2overall_coeff_var=L2.overall_coeff_var
> gen double L2overall_coeff_var_sq=L2.overall_coeff_var_sq
> gen double L3overall_coeff_var=L3.overall_coeff_var
> gen double L3overall_coeff_var_sq=L3.overall_coeff_var_sq
>
> and use these in the estimation command.
>
> Internally, to implement the FD estimator xtivreg2 using the D time-series operator, and it might not be handling the case of D combined with L properly.
>
> Cheers,
> Mark
>
>> -----Original Message-----
>> From: [email protected]
>> [mailto:[email protected]] On Behalf Of
>> Nils Braakmann
>> Sent: Thursday, December 17, 2009 1:02 PM
>> To: [email protected]
>> Subject: Re: st: AW: Problem with xtivreg2
>>
>> Nope. In fact, I would like to avoid having to do this if possible.
>> The code runs in a research data center, which means that all output
>> has to be checked for confidentiality, and I am not sure whether they
>> would allow me to use -trace- (using it obviously increases their work
>> load by a lot). I'll try to recreate the error using data I have on my
>> pc, but I'm not sure if this works. I basically hoped for an easy
>> solution in the spirit of "Your variable name is too long", but if
>> this problem turns out to be more difficult, I'll have to think of
>> something else...
>>
>> Best,
>> Nils
>>
>> On Thu, Dec 17, 2009 at 1:36 PM, Martin Weiss
>> <[email protected]> wrote:
>> >
>> > <>
>> >
>> > Have you -trace-d the error yet?
>> >
>> >
>> >
>> > HTH
>> > Martin
>> >
>> >
>> > -----Ursprüngliche Nachricht-----
>> > Von: [email protected]
>> > [mailto:[email protected]] Im Auftrag
>> von Nils Braakmann
>> > Gesendet: Donnerstag, 17. Dezember 2009 13:27
>> > An: [email protected]
>> > Betreff: st: Problem with xtivreg2
>> >
>> > Dear all,
>> >
>> > I am just revisiting an old paper (running under version
>> control for
>> > version 9.2 under a version 10.2 maching in a research data center)
>> > and encounter a strange problem with xtivreg2.
>> >
>> > In february I ran the following code:
>> >
>> > xtivreg2 `depvar' `indepvars' `exovars' (overall_coeff_var
>> > overall_coeff_var_sq = L.overall_coeff_var L.overall_coeff_var_sq
>> > L2.overall_coeff_var L2.overall_coeff_var_sq), fd gmm robust
>> >
>> > and got the following results:
>> >
>> > Warning - collinearities detected
>> > Vars dropped: D.year_d7 D.year_d6 D.year_d5
>> >
>> > FIRST DIFFERENCES ESTIMATION
>> > ----------------------------
>> > Number of groups = 1022 Obs per group: min =
>> > 1
>> > avg =
>> > 2.6
>> > max =
>> > 7
>> >
>> > 2-Step GMM estimation
>> > ---------------------
>> >
>> > Statistics robust to heteroskedasticity
>> >
>> > Number of obs =
>> > 2627
>> > F( 24, 2602) =
>> > 4.62
>> > Prob > F =
>> > 0.0000
>> > Total (centered) SS = 113.3549125
>> Centered R2 =
>> > -0.0026
>> > Total (uncentered) SS = 113.83846
>> Uncentered R2 =
>> > 0.0017
>> > Residual SS = 113.6464909 Root
>> MSE =
>> > .208
>> >
>> >
>> --------------------------------------------------------------
>> --------------
>> > --
>> > | Robust
>> > D. |
>> > log_return~d | Coef. Std. Err. z P>|z| [95% Conf.
>> > Interval]
>> >
>> -------------+------------------------------------------------
>> --------------
>> > --
>> > overall_co~q |
>> > D1. | 6.448465 5.537348 1.16 0.244 -4.404538
>> > 17.30147
>> > overall_co~r |
>> > D1. | -4.672704 3.66052 -1.28 0.202 -11.84719
>> > 2.501784
>> > year_d13 |
>> > D1. | .1064509 .1903947 0.56 0.576 -.2667159
>> > .4796177
>> > year_d12 |
>> > D1. | .0660634 .1644337 0.40 0.688 -.2562208
>> > .3883475
>> > year_d11 |
>> > D1. | .085434 .1305051 0.65 0.513 -.1703513
>> > .3412193
>> > year_d10 |
>> > D1. | .0852973 .0898904 0.95 0.343 -.0908845
>> > .2614792
>> > year_d9 |
>> > D1. | .053293 .0615972 0.87 0.387 -.0674353
>> > .1740213
>> > year_d8 |
>> > D1. | .0145953 .0313659 0.47 0.642 -.0468806
>> > .0760713
>> > work_council |
>> > D1. | .0044819 .0428912 0.10 0.917 -.0795833
>> > .0885471
>> > bargain_firm |
>> > D1. | -.0117102 .022009 -0.53 0.595 -.054847
>> > .0314266
>> > bargain_br~h |
>> > D1. | .0176108 .0160317 1.10 0.272 -.0138109
>> > .0490324
>> > share_women |
>> > D1. | -.3123268 .3655674 -0.85 0.393 -1.028826
>> > .4041721
>> > invest_head |
>> > D1. | 3.60e-07 3.37e-07 1.07 0.286 -3.02e-07
>> > 1.02e-06
>> > share_acad~c |
>> > D1. | -.1039652 .6003136 -0.17 0.863 -1.280558
>> > 1.072628
>> > share_other |
>> > D1. | -.6331339 .5519297 -1.15 0.251 -1.714896
>> > .4486285
>> > share_whit~l |
>> > D1. | .2077333 .4054792 0.51 0.608 -.5869912
>> > 1.002458
>> > share_unsk~d |
>> > D1. | -.5069874 .2573503 -1.97 0.049 -1.011385
>> > -.00259
>> > share_azubi |
>> > D1. | -.4534471 1.215144 -0.37 0.709 -2.835085
>> > 1.928191
>> > share_age50 |
>> > D1. | .1681527 .3965507 0.42 0.672 -.6090725
>> > .9453778
>> > share_a~4050 |
>> > D1. | .2137222 .3380754 0.63 0.527 -.4488933
>> > .8763378
>> > share_age30 |
>> > D1. | -.3064232 .4795452 -0.64 0.523 -1.246315
>> > .6334681
>> > firmsize_sq |
>> > D1. | 3.84e-08 1.06e-08 3.61 0.000 1.76e-08
>> > 5.91e-08
>> > firmsize |
>> > D1. | -.0005478 .0001474 -3.72 0.000 -.0008367
>> > -.0002589
>> > avg_wage |
>> > D1. | .0032925 .0037199 0.89 0.376 -.0039983
>> > .0105834
>> > _cons | -.003982 .0268464 -0.15 0.882 -.0566
>> > .048636
>> >
>> --------------------------------------------------------------
>> --------------
>> > --
>> > Anderson canon. corr. LR statistic (underidentification test):
>> > 108.788
>> > Chi-sq(3) P-val =
>> > 0.0000
>> > Test statistic(s) not robust
>> >
>> --------------------------------------------------------------
>> --------------
>> > --
>> > Cragg-Donald F statistic (weak identification test):
>> > 27.483
>> > Stock-Yogo weak ID test critical values: 5% maximal IV
>> relative bias
>> > 11.04
>> > 10% maximal IV relative bias
>> > 7.56
>> > 20% maximal IV relative bias
>> > 5.57
>> > 30% maximal IV relative bias
>> > 4.73
>> > 10% maximal IV size
>> > 16.87
>> > 15% maximal IV size
>> > 9.93
>> > 20% maximal IV size
>> > 7.54
>> > 25% maximal IV size
>> > 6.28
>> > Test statistic(s) not robust
>> > Source: Stock-Yogo (2005). Reproduced by permission.
>> >
>> --------------------------------------------------------------
>> --------------
>> > --
>> > Hansen J statistic (overidentification test of all instruments):
>> > 1.761
>> > Chi-sq(2) P-val =
>> > 0.4147
>> >
>> --------------------------------------------------------------
>> --------------
>> > --
>> > Instrumented: D.overall_coeff_var_sq D.overall_coeff_var
>> > Included instruments: D.year_d13 D.year_d12 D.year_d11
>> D.year_d10 D.year_d9
>> > D.year_d8 D.work_council D.bargain_firm
>> > D.bargain_branch
>> > D.share_women D.invest_head D.share_academic
>> > D.share_other
>> > D.share_whitecol D.share_unskilled
>> D.share_azubi
>> > D.share_age50 D.share_age4050 D.share_age30
>> > D.firmsize_sq
>> > D.firmsize D.avg_wage
>> > Excluded instruments: L2D.overall_coeff_var_sq L2D.overall_coeff_var
>> > LD.overall_coeff_var_sq LD.overall_coeff_var
>> > Dropped collinear: D.year_d7 D.year_d6 D.year_d5
>> >
>> --------------------------------------------------------------
>> --------------
>> > --
>> >
>> >
>> >
>> > Some days ago, I changed the instruments from lagged first
>> and second
>> > differences to lagged second and third differences, gmm to 2sls and
>> > added clustered standard errors and the option -first- to
>> have a look
>> > at the first stage results. In other words, I ran the
>> following code:
>> > xtivreg2 `depvar' `indepvars' `exovars' (overall_coeff_var
>> > overall_coeff_var_sq = L2.overall_coeff_var L2.overall_coeff_var_sq
>> > L3.overall_coeff_var L3.overall_coeff_var_sq), fd
>> cluster(idnum) first
>> >
>> > This time I get an error message:
>> >
>> > D.overall_coeff_v invalid name
>> > r(198); t=8.63 11:10:53
>> >
>> > As far as I know neither the Stata Version, nor the dataset
>> used have
>> > been changed in the meantime. SAme applies to the macros
>> used for the
>> > dependent and independent variables.
>> >
>> > I'd greatly appreciate any input on this issue as I am a bit (well,
>> > really) lost at present.
>> >
>> > Best,
>> > Nils
>> > *
>> > * 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/
>>
>
>
> --
> Heriot-Watt University is a Scottish charity
> registered under charity number SC000278.
>
>
> *
> * 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/
