No standard errors reported for some variables when using xtlsdvc

[Ona Ciocyte]

Hello,

I am trying to use xtlsdvc to correct for the bias in the dynamic panel model.

However, I do not get the standard errors for some of the variables. If I exclude one of the 2 variables, I still do not get the standard errors for the remaining one. What can be a problem?

-----------------------------------------------------------------------------------------------------------------------------------------------------------

. xtlsdvc cab gap infl Ldebt Lmt el gov Lcab_dEU gap_dEU infl_dEU Ldebt_dEU el_dEU gov_dEU dEU td19
> 93-td1997 td1993_dEU-td1997_dEU ///
> if year>=1992 & year<=1997, initial(bb) bias(1) vcov(100)
Note: Bias correction initialized by Blundell and Bond estimator

Note: Blundell and Bond estimator is implemented through
the user-written Stata command -xtabond2- by David Roodman,
Center for Global Development, Washington, DC [email protected]
note: Bias correction up to order O(1/T)

LSDVC dynamic regression
(bootstrapped SE)

------------------------------------------------------------------------------
cab | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
cab |
L1. | 1.185347 .0518106 22.88 0.000 1.0838 1.28689
gap | -.3446312 .36032 -0.96 0.339 -1.050846 .361583
infl | .5092328 .2976677 1.71 0.087 -.0741852 1.092651
Ldebt | .0860462 .0610081 1.41 0.158 -.0335274 .2056199
Lmt | 1.018077 .9867556 1.03 0.302 -.915928 2.952083
el | -.259387 .5843941 -0.44 0.657 -1.404778 .8860044
gov | 2.405338 6.46627 0.37 0.710 -10.26832 15.07899
Lcab_dEU | -.9366056 .3216365 -2.91 0.004 -1.567002 -.3062096
gap_dEU | .2761265 .5098341 0.54 0.588 -.72313 1.275383
infl_dEU | -.1643232 .510955 -0.32 0.748 -1.165777 .8371301
Ldebt_dEU | -.0136853 .0566955 -0.24 0.809 -.1248065 .0974358
el_dEU | .3343402 .7964652 0.42 0.675 -1.226703 1.895383
gov_dEU | -3.139014 7.736398 -0.41 0.685 -18.30208 12.02405
dEU | 5.214859 6.621693 0.79 0.431 -7.76342 18.19314
td1993 | .4964171 .9872596 0.50 0.615 -1.438576 2.43141
td1994 | 1.508747 .971539 1.55 0.120 -.3954347 3.412928
td1995 | 1.453708 1.298864 1.12 0.263 -1.092019 3.999434
td1996 | .67081 1.363475 0.49 0.623 -2.001552 3.343172
td1997 | .9267471 1.436042 0.65 0.519 -1.887844 3.741339
td1993_dEU | -.436194 1.426415 -0.31 0.760 -3.231917 2.359529
td1994_dEU | -2.460774 1.684007 -1.46 0.144 -5.761368 .8398204
td1995_dEU | -3.486547 1.921637 -1.81 0.070 -7.252887 .2797924
td1996_dEU | -.311034 . . . . .
td1997_dEU | .0798837 . . . . .
------------------------------------------------------------------------------

-----------------------------------------------------------------------------------------------------------------------------


Thanks,

Ona

[Nick Cox]

Your results don't include information on the number of observations used. Usually, on this kind of information, I guess at overfitting, too many parameters chasing too little data.

It's hard to read your results (please read and act on FAQ Advice #12 which explains how to display them as CODE) but there seem many unimportant predictors. You may well have a rationale there, but that is a danger signal.

[Philipp Allroggen]

You have to use the -vcov- option, this will give you standard errors via bootstrap.