Hi Stata gooroos!
I am trying to figure out what is it that xtreg does differently from xtscc that affects estimation of t-ratios for time variables. I have a data set on firms over time observing their inventory sizes (in days). Now if I put in a time fixed effect it doesn't report t-stats in xtscc while it does so for xtreg. I am hoping someone can help
*summary of the data*
. sum avginventorydays yr1-yr14
Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
avginvento~s | 812 71.52417 42.35176 4.75 345.3
yr1 | 812 .0714286 .2576981 0 1
yr2 | 812 .0714286 .2576981 0 1
yr3 | 812 .0714286 .2576981 0 1
yr4 | 812 .0714286 .2576981 0 1
-------------+--------------------------------------------------------
yr5 | 812 .0714286 .2576981 0 1
yr6 | 812 .0714286 .2576981 0 1
yr7 | 812 .0714286 .2576981 0 1
yr8 | 812 .0714286 .2576981 0 1
yr9 | 812 .0714286 .2576981 0 1
-------------+--------------------------------------------------------
yr10 | 812 .0714286 .2576981 0 1
yr11 | 812 .0714286 .2576981 0 1
yr12 | 812 .0714286 .2576981 0 1
yr13 | 812 .0714286 .2576981 0 1
yr14 | 812 .0714286 .2576981 0 1
*xtreg regression*
. xtreg avginventorydays yr1-yr13, fe
Fixed-effects (within) regression Number of obs = 812
Group variable: n_company Number of groups = 58
R-sq: within = 0.2527 Obs per group: min = 14
between = 0.0000 avg = 14.0
overall = 0.0963 max = 14
F(13,741) = 19.27
corr(u_i, Xb) = 0.0000 Prob> F = 0.0000
------------------------------------------------------------------------------
avginvento~s | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
yr1 | 44.38966 4.390504 10.11 0.000 35.77035 53.00896
yr2 | 43.56345 4.390504 9.92 0.000 34.94414 52.18276
yr3 | 35.44948 4.390504 8.07 0.000 26.83017 44.06879
yr4 | 25.79155 4.390504 5.87 0.000 17.17224 34.41086
yr5 | 22.10862 4.390504 5.04 0.000 13.48931 30.72793
yr6 | 18.50586 4.390504 4.21 0.000 9.886554 27.12517
yr7 | 27.1569 4.390504 6.19 0.000 18.53759 35.77621
yr8 | 20.4119 4.390504 4.65 0.000 11.79259 29.03121
yr9 | 16.31241 4.390504 3.72 0.000 7.693105 24.93172
yr10 | 16.02845 4.390504 3.65 0.000 7.409139 24.64776
yr11 | 16.16276 4.390504 3.68 0.000 7.54345 24.78207
yr12 | 8.121896 4.390504 1.85 0.065 -.4974127 16.74121
yr13 | .8931034 4.390504 0.20 0.839 -7.726206 9.512412
_cons | 50.46017 3.104555 16.25 0.000 44.3654 56.55494
-------------+----------------------------------------------------------------
sigma_u | 33.590036
sigma_e | 23.64359
rho | .66869195 (fraction of variance due to u_i)
------------------------------------------------------------------------------
F test that all u_i=0: F(57, 741) = 28.26 Prob> F = 0.0000
*xtscc*
xtscc avginventorydays yr1-yr13, fe
Regression with Driscoll-Kraay standard errors Number of obs = 812
Method: Fixed-effects regression Number of groups = 58
Group variable (i): n_company F( 13, 57) = .
maximum lag: 2 Prob> F = .
within R-squared = 0.2527
------------------------------------------------------------------------------
| Drisc/Kraay
avginvento~s | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
yr1 | 44.38966 4.19e-12 . 0.000 44.38966 44.38966
yr2 | 43.56345 4.19e-12 . 0.000 43.56345 43.56345
yr3 | 35.44948 4.19e-12 . 0.000 35.44948 35.44948
yr4 | 25.79155 4.19e-12 . 0.000 25.79155 25.79155
yr5 | 22.10862 4.19e-12 . 0.000 22.10862 22.10862
yr6 | 18.50586 4.19e-12 . 0.000 18.50586 18.50586
yr7 | 27.1569 4.19e-12 . 0.000 27.1569 27.1569
yr8 | 20.4119 4.19e-12 . 0.000 20.4119 20.4119
yr9 | 16.31241 4.19e-12 . 0.000 16.31241 16.31241
yr10 | 16.02845 4.19e-12 . 0.000 16.02845 16.02845
yr11 | 16.16276 4.20e-12 . 0.000 16.16276 16.16276
yr12 | 8.121896 4.22e-12 . 0.000 8.121896 8.121896
yr13 | .8931034 4.35e-12 . 0.000 .8931034 .8931034
_cons | 50.46017 4.19e-12 . 0.000 50.46017 50.46017
------------------------------------------------------------------------------
I'd like to know what to infer from this ... how is it that the std. errors have shrunk so much! Any help would be great.
Alex
________________________________
Now you can invite friends from Facebook and other groups to join you on Windows Live™ Messenger. Add them now!
_________________________________________________________________
Instantly invite friends from Facebook and other social networks to join you on Windows Live™ Messenger.
https://www.invite2messenger.net/im/?source=TXT_EML_WLH_InviteFriends
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
