Random-Effects Model: Sigma u equals zero, Breusch-Pagan Lagrange Multiplier equals 1
(This is my first time posting something on StataList, so if my formatting or explanations are wrong or missing elements, please let me know, and I will make any changes accordingly.)
I am using Stata 15
I have recently encountered a problem while running my Random-Effects Model.
I am using an unbalanced Panel Data Set which consists of time-invariant and time-variant variables. The time-invariant variables are cultural dimensions and the time-variant variables are economic, population and infrastructure ones (GDP, Gini coefffient, foreign population etc.) My dependent variable is TTT_tot. My control variables are NE, _ios and _gp.
I have N=20 groups/observational units in my sample, and T=64.
I have calculated three models with the following do.file-code:
I want to compare a standard OLS regression, the Random-Effects Model and a Fixed-Effects model with one another, to be able to compare their standard errors and sigma u and sigma e values. However, I constantly receive a 0 for my sigma u value in my Random-Effects model.
Here we can see the 0 value for sigma_u. Normally this would mean that the RE-model is degenerated. But this seems weird, since the RE model results in more statistically significant results, with better standard errors compared to my OLS regression results.
I have furthermore conducted a Breusch and Pagan Langrangian Multiplier test to determine whether I can even use a Random-Effects Model. My Prob > chibar2 equals 1, which seems peculiar to me.
All the best,
Johanna
I am using Stata 15
I have recently encountered a problem while running my Random-Effects Model.
I am using an unbalanced Panel Data Set which consists of time-invariant and time-variant variables. The time-invariant variables are cultural dimensions and the time-variant variables are economic, population and infrastructure ones (GDP, Gini coefffient, foreign population etc.) My dependent variable is TTT_tot. My control variables are NE, _ios and _gp.
I have N=20 groups/observational units in my sample, and T=64.
I have calculated three models with the following do.file-code:
Code:
reg TTT_tot `eco_list' `cul_list' `infr_list' `pop_list' NE _ios _gp, vce(robust) xtset CountryCode xtreg TTT_tot `eco_list' `cul_list' `infr_list' `pop_list' NE _ios _gp ,fe vce(robust) xtreg TTT_tot `eco_list' `cul_list' `infr_list' `pop_list' NE _ios _gp ,re vce(robust)
Code:
Random-effects GLS regression Number of obs = 11,934
Group variable: CountryCode Number of groups = 19
R-sq: Obs per group:
within = 0.0541 min = 594
between = 1.0000 avg = 628.1
overall = 0.1528 max = 630
Wald chi2(13) = .
corr(u_i, X) = 0 (assumed) Prob > chi2 = .
(Std. Err. adjusted for 19 clusters in CountryCode)
------------------------------------------------------------------------------------------------
| Robust
TTT_tot | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------------------------+----------------------------------------------------------------
GDP | 2.53e-06 4.09e-06 0.62 0.536 -5.48e-06 .0000105
Gini | -.0559253 .5105896 -0.11 0.913 -1.056663 .9448119
net_national_income | -.001845 .0017765 -1.04 0.299 -.005327 .0016369
UncertaintyAvoidanceSP | -2.461559 .2853507 -8.63 0.000 -3.020836 -1.902281
FutureOrientationSP | -4.779695 .704769 -6.78 0.000 -6.161016 -3.398373
PowerDistanceSP | -5.717074 .2371044 -24.11 0.000 -6.18179 -5.252358
CollectivismISPInstitutional | 17.74914 .52927 33.54 0.000 16.71179 18.78649
HumaneOrientationSP | -6.138548 .0818562 -74.99 0.000 -6.298983 -5.978113
PerformanceOrientationSP | -7.951868 .4913199 -16.18 0.000 -8.914837 -6.988899
CollectivismIISPIngroupCol | -1.273744 .7499475 -1.70 0.089 -2.743615 .1961257
GenderEgalitarianismSP | .5030384 .5760547 0.87 0.383 -.6260081 1.632085
AssertivenessSP | 11.80073 .3703147 31.87 0.000 11.07492 12.52653
UncertaintyAvoidanceSV | -9.892615 .6663491 -14.85 0.000 -11.19864 -8.586595
FutureOrientationSV | 1.941818 .2248855 8.63 0.000 1.501051 2.382586
PowerDistanceSV | 11.56595 .7878502 14.68 0.000 10.02179 13.11011
CollectivismISVInstitutional | 19.87028 .9030751 22.00 0.000 18.10028 21.64027
HumaneOrientationSV | 8.966052 .961623 9.32 0.000 7.081306 10.8508
PerformanceOrientationSV | -2.439227 .112894 -21.61 0.000 -2.660495 -2.217959
CollectivismIISVIngroupCol | 1.623323 .7085817 2.29 0.022 .2345288 3.012118
GenderEgalitarianismSV | -8.887995 .3053794 -29.10 0.000 -9.486528 -8.289462
AssertivenessSV | 8.362968 .623689 13.41 0.000 7.14056 9.585376
Broadband_subscription | .0118887 .0081043 1.47 0.142 -.0039954 .0277728
mobile_broadband_subscriptions | -.0003924 .0004245 -0.92 0.355 -.0012243 .0004396
urban_pop | .5537966 1.768498 0.31 0.754 -2.912395 4.019988
new_foreign_pop_ | .2852685 .9709759 0.29 0.769 -1.617809 2.188346
new_foreign_pop_growth | -.0539973 .1212795 -0.45 0.656 -.2917008 .1837061
edu_post_sec | 2.001969 1.610598 1.24 0.214 -1.154745 5.158683
edu_tert | .2519419 .7611614 0.33 0.741 -1.239907 1.743791
NE | -6.031001 1.565617 -3.85 0.000 -9.099554 -2.962448
_ios | .0218838 .0709499 0.31 0.758 -.1171756 .1609431
_gp | .0341277 .0900629 0.38 0.705 -.1423923 .2106476
_cons | -102.1462 12.21292 -8.36 0.000 -126.0831 -78.20936
-------------------------------+----------------------------------------------------------------
sigma_u | 0
sigma_e | 12.622343
rho | 0 (fraction of variance due to u_i)
------------------------------------------------------------------------------------------------
I have furthermore conducted a Breusch and Pagan Langrangian Multiplier test to determine whether I can even use a Random-Effects Model. My Prob > chibar2 equals 1, which seems peculiar to me.
All the best,
Johanna
Comment