Hi All,
I am running a regression with dependant variable happiness and independent variables are various individual-level and country-level variables. I have 30 countries and due to collinearity with the macro variables, I have grouped the countries into regions. I also cluster the standard errors at country level. After the regression I ran linktest and estat ovtest and they are significant (p-value = 0.000), suggesting that the null hypothesis of no omitted variable or misspecification can be rejected. I am a bit surprised with the results as I have used the quite similar model on one country but only individual-level variables and it was alright according to these tests. I am not sure whether there may be a mistake with the way I clustered the standard errors? Or these test are not appropriate when I have clustered errors? Please if anyone can give any feedback.
. regress HappinessTenGroups female Age agesq Divorced Widowed Single HaveChildren Urban Secondary Tertiar
> y ShortUnemployed LongUnemployed Retired Homemaker Student Other Good Fair Bad Verybad SecondQuartile Thr
> idQuartile Fourthquartile IncomeNotReported GDPcapitaLOG GDPgrowthannual Unemploymenttotaloftotal Inflati
> onconsumerpricesannu ControlofCorruptionEstimate VoiceandAccountability PeopleTrust EU15 A10 NewEntrants
> if Wave==2, cluster (Y11_Country)
Linear regression Number of obs = 33,696
F(28, 29) = .
Prob > F = .
R-squared = 0.2665
Root MSE = 1.6489
(Std. Err. adjusted for 30 clusters in Y11_Country)
Robust
HappinessTenGroups Coef. Std. Err. t P>t [95% Conf. Interval]
female .1166555 .0211694 5.51 0.000 .0733592 .1599519
Age -.0472745 .0048993 -9.65 0.000 -.0572947 -.0372543
agesq .0004908 .0000489 10.04 0.000 .0003907 .0005908
Divorced -.8107275 .0615956 -13.16 0.000 -.9367046 -.6847504
Widowed -.8226137 .0428381 -19.20 0.000 -.9102274 -.7350001
Single -.5200787 .0459098 -11.33 0.000 -.6139747 -.4261827
HaveChildren .057797 .0145225 3.98 0.000 .0280952 .0874989
Urban -.0081242 .0287305 -0.28 0.779 -.0668848 .0506363
Secondary .1397538 .0602732 2.32 0.028 .0164813 .2630263
Tertiary .2091507 .0704919 2.97 0.006 .0649786 .3533227
ShortUnemployed -.4204632 .087762 -4.79 0.000 -.5999567 -.2409696
LongUnemployed -.5808111 .1199096 -4.84 0.000 -.8260537 -.3355684
Retired .0994038 .0496802 2.00 0.055 -.0022036 .2010112
Homemaker .0125494 .0591562 0.21 0.833 -.1084386 .1335375
Student .2840124 .0571309 4.97 0.000 .1671666 .4008582
Other -.0318354 .0604765 -0.53 0.603 -.1555237 .0918529
Good -.4908626 .0334783 -14.66 0.000 -.5593334 -.4223917
Fair -.9904155 .0477664 -20.73 0.000 -1.088109 -.8927222
Bad -1.875238 .061113 -30.68 0.000 -2.000228 -1.750248
Verybad -2.940311 .122485 -24.01 0.000 -3.190821 -2.689801
SecondQuartile .3019176 .0483881 6.24 0.000 .2029528 .4008823
ThridQuartile .395378 .0615345 6.43 0.000 .2695257 .5212302
Fourthquartile .547845 .0756035 7.25 0.000 .3932185 .7024716
IncomeNotReported .3191297 .051822 6.16 0.000 .2131417 .4251176
GDPcapitaLOG .1208351 .1860239 0.65 0.521 -.2596266 .5012968
GDPgrowthannual .0414192 .0205828 2.01 0.054 -.0006773 .0835158
Unemploymenttotaloftotal -.0163014 .0078012 -2.09 0.046 -.0322566 -.0003461
Inflationconsumerpricesannu -.0174284 .0239972 -0.73 0.473 -.0665082 .0316514
ControlofCorruptionEstimate .2523979 .1193416 2.11 0.043 .008317 .4964788
VoiceandAccountability .3545206 .4847643 0.73 0.470 -.6369338 1.345975
PeopleTrust .0980505 .0082492 11.89 0.000 .0811791 .114922
EU15 -.2124707 .2893856 -0.73 0.469 -.8043308 .3793894
A10 .0814843 .317972 0.26 0.800 -.5688414 .73181
NewEntrants -.5619843 .4754771 -1.18 0.247 -1.534444 .4104755
_cons 6.501403 1.767906 3.68 0.001 2.885629 10.11718
. linktest
Source SS df MS Number of obs = 33,696
F(2, 33693) = 6176.89
Model 33472.3736 2 16736.1868 Prob > F = 0.0000
Residual 91290.6006 33,693 2.7094827 R-squared = 0.2683
Adj R-squared = 0.2682
Total 124762.974 33,695 3.70271477 Root MSE = 1.6461
HappinessT~s Coef. Std. Err. t P>t [95% Conf. Interval]
_hat 1.789596 .0870475 20.56 0.000 1.61898 1.960212
_hatsq -.0565058 .0061958 -9.12 0.000 -.0686497 -.0443618
_cons -2.695989 .3030599 -8.90 0.000 -3.289997 -2.101981
estat ovtest
Ramsey RESET test using powers of the fitted values of HappinessTenGroups
Ho: model has no omitted variables
F(3, 33658) = 54.44
Prob > F = 0.0000
Many thanks,
Mirjana
I am running a regression with dependant variable happiness and independent variables are various individual-level and country-level variables. I have 30 countries and due to collinearity with the macro variables, I have grouped the countries into regions. I also cluster the standard errors at country level. After the regression I ran linktest and estat ovtest and they are significant (p-value = 0.000), suggesting that the null hypothesis of no omitted variable or misspecification can be rejected. I am a bit surprised with the results as I have used the quite similar model on one country but only individual-level variables and it was alright according to these tests. I am not sure whether there may be a mistake with the way I clustered the standard errors? Or these test are not appropriate when I have clustered errors? Please if anyone can give any feedback.
. regress HappinessTenGroups female Age agesq Divorced Widowed Single HaveChildren Urban Secondary Tertiar
> y ShortUnemployed LongUnemployed Retired Homemaker Student Other Good Fair Bad Verybad SecondQuartile Thr
> idQuartile Fourthquartile IncomeNotReported GDPcapitaLOG GDPgrowthannual Unemploymenttotaloftotal Inflati
> onconsumerpricesannu ControlofCorruptionEstimate VoiceandAccountability PeopleTrust EU15 A10 NewEntrants
> if Wave==2, cluster (Y11_Country)
Linear regression Number of obs = 33,696
F(28, 29) = .
Prob > F = .
R-squared = 0.2665
Root MSE = 1.6489
(Std. Err. adjusted for 30 clusters in Y11_Country)
Robust
HappinessTenGroups Coef. Std. Err. t P>t [95% Conf. Interval]
female .1166555 .0211694 5.51 0.000 .0733592 .1599519
Age -.0472745 .0048993 -9.65 0.000 -.0572947 -.0372543
agesq .0004908 .0000489 10.04 0.000 .0003907 .0005908
Divorced -.8107275 .0615956 -13.16 0.000 -.9367046 -.6847504
Widowed -.8226137 .0428381 -19.20 0.000 -.9102274 -.7350001
Single -.5200787 .0459098 -11.33 0.000 -.6139747 -.4261827
HaveChildren .057797 .0145225 3.98 0.000 .0280952 .0874989
Urban -.0081242 .0287305 -0.28 0.779 -.0668848 .0506363
Secondary .1397538 .0602732 2.32 0.028 .0164813 .2630263
Tertiary .2091507 .0704919 2.97 0.006 .0649786 .3533227
ShortUnemployed -.4204632 .087762 -4.79 0.000 -.5999567 -.2409696
LongUnemployed -.5808111 .1199096 -4.84 0.000 -.8260537 -.3355684
Retired .0994038 .0496802 2.00 0.055 -.0022036 .2010112
Homemaker .0125494 .0591562 0.21 0.833 -.1084386 .1335375
Student .2840124 .0571309 4.97 0.000 .1671666 .4008582
Other -.0318354 .0604765 -0.53 0.603 -.1555237 .0918529
Good -.4908626 .0334783 -14.66 0.000 -.5593334 -.4223917
Fair -.9904155 .0477664 -20.73 0.000 -1.088109 -.8927222
Bad -1.875238 .061113 -30.68 0.000 -2.000228 -1.750248
Verybad -2.940311 .122485 -24.01 0.000 -3.190821 -2.689801
SecondQuartile .3019176 .0483881 6.24 0.000 .2029528 .4008823
ThridQuartile .395378 .0615345 6.43 0.000 .2695257 .5212302
Fourthquartile .547845 .0756035 7.25 0.000 .3932185 .7024716
IncomeNotReported .3191297 .051822 6.16 0.000 .2131417 .4251176
GDPcapitaLOG .1208351 .1860239 0.65 0.521 -.2596266 .5012968
GDPgrowthannual .0414192 .0205828 2.01 0.054 -.0006773 .0835158
Unemploymenttotaloftotal -.0163014 .0078012 -2.09 0.046 -.0322566 -.0003461
Inflationconsumerpricesannu -.0174284 .0239972 -0.73 0.473 -.0665082 .0316514
ControlofCorruptionEstimate .2523979 .1193416 2.11 0.043 .008317 .4964788
VoiceandAccountability .3545206 .4847643 0.73 0.470 -.6369338 1.345975
PeopleTrust .0980505 .0082492 11.89 0.000 .0811791 .114922
EU15 -.2124707 .2893856 -0.73 0.469 -.8043308 .3793894
A10 .0814843 .317972 0.26 0.800 -.5688414 .73181
NewEntrants -.5619843 .4754771 -1.18 0.247 -1.534444 .4104755
_cons 6.501403 1.767906 3.68 0.001 2.885629 10.11718
. linktest
Source SS df MS Number of obs = 33,696
F(2, 33693) = 6176.89
Model 33472.3736 2 16736.1868 Prob > F = 0.0000
Residual 91290.6006 33,693 2.7094827 R-squared = 0.2683
Adj R-squared = 0.2682
Total 124762.974 33,695 3.70271477 Root MSE = 1.6461
HappinessT~s Coef. Std. Err. t P>t [95% Conf. Interval]
_hat 1.789596 .0870475 20.56 0.000 1.61898 1.960212
_hatsq -.0565058 .0061958 -9.12 0.000 -.0686497 -.0443618
_cons -2.695989 .3030599 -8.90 0.000 -3.289997 -2.101981
estat ovtest
Ramsey RESET test using powers of the fitted values of HappinessTenGroups
Ho: model has no omitted variables
F(3, 33658) = 54.44
Prob > F = 0.0000
Many thanks,
Mirjana
