Dear STATA Community,
i am estimating following model using panel data with dummy control variables like religion colonial legacy and country fixed effects. i have one endogenous regressor and instrument variable. Although the results include weak instrument test but it does not contain the endogeniety test and i am also confused by warning message "Warning: estimated covariance matrix......" therefore as mentioned in the ivre2 manual i reestimated the model with partial (varlist) option..now i have all the post-estimation results but the coefficients for dummy variables which are important control variables are not given in the coefficient table ...i dont understand what happens with these control variables and how can i interpret results for dropped control variables in the partial out model.....
as mentioned in the ivreg2 help manual for the partial out option it says " By the Frisch-Waugh-Lovell (FWL) theorem, in IV, two-step GMM and LIML estimation the coefficients for the remaining regressors are the same as those that would be obtained if the variables were not partialled out."
does it mean that the coefficient will not be changed for those that are dropped in the partial out model... can i merge the results for instance the coefficients table from the first model and postestimation results from the partial out model and then provide a note "post-estimation statistics are obtained using Frisch-Waugh-Lovell Theorem to partial out the exogenous regressors"
i am estimating following model using panel data with dummy control variables like religion colonial legacy and country fixed effects. i have one endogenous regressor and instrument variable. Although the results include weak instrument test but it does not contain the endogeniety test and i am also confused by warning message "Warning: estimated covariance matrix......" therefore as mentioned in the ivre2 manual i reestimated the model with partial (varlist) option..now i have all the post-estimation results but the coefficients for dummy variables which are important control variables are not given in the coefficient table ...i dont understand what happens with these control variables and how can i interpret results for dropped control variables in the partial out model.....
Code:
ivreg2 lninval (bm=msphund) lnpn polity2 popd religion i.colonyg i.ccode, cluster(ccode) endog(bm)
Code:
Warning - collinearities detected
Vars dropped: 14.ccode 16.ccode 33.ccode 36.ccode 40.ccode 43.ccode
IV (2SLS) estimation
--------------------
Estimates efficient for homoskedasticity only
Statistics robust to heteroskedasticity and clustering on ccode
Number of clusters (ccode) = 42 Number of obs = 1176
F( 45, 41) = 4.02
Prob > F = 0.0000
Total (centered) SS = 8782.392021 Centered R2 = 0.9527
Total (uncentered) SS = 102392.8999 Uncentered R2 = 0.9959
Residual SS = 415.5114745 Root MSE = .5944
------------------------------------------------------------------------------
| Robust
lninval | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
bm | -.0550026 .0148295 -3.71 0.000 -.0840679 -.0259372
lnpn | .6344185 .1019784 6.22 0.000 .4345445 .8342925
polity2 | -.0221709 .0163135 -1.36 0.174 -.0541447 .0098029
popd | -.000797 .0045606 -0.17 0.861 -.0097356 .0081416
religion | -1.545679 .655505 -2.36 0.018 -2.830446 -.2609131
|
colonyg |
France | .6188915 1.295867 0.48 0.633 -1.92096 3.158743
GB | 1.185666 1.627479 0.73 0.466 -2.004134 4.375467
Italy | .8160768 1.44639 0.56 0.573 -2.018796 3.65095
Portugal | .8450834 1.63738 0.52 0.606 -2.364123 4.054289
Spain | -1.075271 1.421021 -0.76 0.449 -3.860421 1.70988
|
ccode |
2 | -.1374123 .2565583 -0.54 0.592 -.6402574 .3654327
3 | -.1817275 .3549695 -0.51 0.609 -.877455 .5139999
4 | .4715192 .7613694 0.62 0.536 -1.020737 1.963776
5 | -.344699 .3413542 -1.01 0.313 -1.013741 .324343
6 | .9443374 .6825825 1.38 0.167 -.3934998 2.282175
7 | -.3731629 .2848706 -1.31 0.190 -.931499 .1851732
8 | -.9599937 .3675614 -2.61 0.009 -1.680401 -.2395866
9 | .4102447 .7871616 0.52 0.602 -1.132564 1.953053
10 | -2.560205 .9727601 -2.63 0.008 -4.46678 -.6536305
11 | -.4878504 1.429966 -0.34 0.733 -3.290532 2.314831
12 | -1.076217 .3745986 -2.87 0.004 -1.810417 -.3420175
13 | .09228 .2410625 0.38 0.702 -.3801938 .5647538
14 | 0 (omitted)
15 | -.9769211 .3463366 -2.82 0.005 -1.655728 -.2981138
16 | 0 (omitted)
17 | -1.456316 .3746422 -3.89 0.000 -2.190601 -.7220305
18 | .0666229 .3060478 0.22 0.828 -.5332199 .6664656
19 | -.5456202 .4413525 -1.24 0.216 -1.410655 .3194149
20 | .6744012 .6622356 1.02 0.309 -.6235566 1.972359
21 | -.7752258 .2783507 -2.79 0.005 -1.320783 -.2296683
22 | .6843882 .1415277 4.84 0.000 .4069989 .9617774
23 | -.0550631 .3391241 -0.16 0.871 -.719734 .6096079
24 | -.588617 .4212288 -1.40 0.162 -1.41421 .2369764
25 | -.5324383 .6261119 -0.85 0.395 -1.759595 .6947185
26 | 1.18508 .8053307 1.47 0.141 -.393339 2.763499
27 | -2.671196 .5901045 -4.53 0.000 -3.82778 -1.514612
28 | 1.662388 2.332419 0.71 0.476 -2.90907 6.233846
29 | .7522117 .2014798 3.73 0.000 .3573185 1.147105
30 | .2789614 .4216059 0.66 0.508 -.5473709 1.105294
31 | .291211 .7230231 0.40 0.687 -1.125888 1.70831
32 | 1.595191 .6150546 2.59 0.009 .3897066 2.800676
33 | 0 (omitted)
34 | -4.751193 1.205532 -3.94 0.000 -7.113993 -2.388393
35 | .9146707 .5859546 1.56 0.119 -.2337793 2.063121
36 | 0 (omitted)
37 | 2.840891 .4285955 6.63 0.000 2.000859 3.680922
39 | -.6249982 .2278374 -2.74 0.006 -1.071551 -.1784452
40 | 0 (omitted)
41 | -.7446223 .6522119 -1.14 0.254 -2.022934 .5336895
42 | -.6195256 .2473664 -2.50 0.012 -1.104355 -.1346965
43 | 0 (omitted)
|
_cons | 3.131583 1.924318 1.63 0.104 -.640011 6.903177
------------------------------------------------------------------------------
Underidentification test (Kleibergen-Paap rk LM statistic): 18.031
Chi-sq(1) P-val = 0.0000
------------------------------------------------------------------------------
Weak identification test (Cragg-Donald Wald F statistic): 228.980
(Kleibergen-Paap rk Wald F statistic): 36.294
Stock-Yogo weak ID test critical values: 10% maximal IV size 16.38
15% maximal IV size 8.96
20% maximal IV size 6.66
25% maximal IV size 5.53
Source: Stock-Yogo (2005). Reproduced by permission.
NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors.
------------------------------------------------------------------------------
Warning: estimated covariance matrix of moment conditions not of full rank.
overidentification statistic not reported, and standard errors and
model tests should be interpreted with caution.
Possible causes:
number of clusters insufficient to calculate robust covariance matrix
singleton dummy variable (dummy with one 1 and N-1 0s or vice versa)
partial option may address problem.
------------------------------------------------------------------------------
Instrumented: bm
Included instruments: lnpn polity2 popd religion 2.colonyg 3.colonyg 4.colonyg
5.colonyg 6.colonyg 2.ccode 3.ccode 4.ccode 5.ccode
6.ccode 7.ccode 8.ccode 9.ccode 10.ccode 11.ccode 12.ccode
13.ccode 15.ccode 17.ccode 18.ccode 19.ccode 20.ccode
21.ccode 22.ccode 23.ccode 24.ccode 25.ccode 26.ccode
27.ccode 28.ccode 29.ccode 30.ccode 31.ccode 32.ccode
34.ccode 35.ccode 37.ccode 39.ccode 41.ccode 42.ccode
Excluded instruments: msphund
Dropped collinear: 14.ccode 16.ccode 33.ccode 36.ccode 40.ccode 43.ccode
------------------------------------------------------------------------------
.
end of do-file
Code:
ivreg2 lninval (bm=msphund) lnpn polity2 popd religion i.colonyg i.ccode, cluster(ccode) endog(bm) partial(religion i.colonyg i.ccode)
Code:
Warning - collinearities detected
Vars dropped: 14.ccode 16.ccode 33.ccode 36.ccode 40.ccode 43.ccode
IV (2SLS) estimation
--------------------
Estimates efficient for homoskedasticity only
Statistics robust to heteroskedasticity and clustering on ccode
Number of clusters (ccode) = 42 Number of obs = 1176
F( 4, 41) = 15.74
Prob > F = 0.0000
Total (centered) SS = 427.3671414 Centered R2 = 0.0277
Total (uncentered) SS = 427.3671414 Uncentered R2 = 0.0277
Residual SS = 415.5114745 Root MSE = .5944
------------------------------------------------------------------------------
| Robust
lninval | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
bm | -.0550026 .0148295 -3.71 0.000 -.0840679 -.0259372
lnpn | .6344185 .1019784 6.22 0.000 .4345445 .8342925
polity2 | -.0221709 .0163135 -1.36 0.174 -.0541447 .0098029
popd | -.000797 .0045606 -0.17 0.861 -.0097356 .0081416
------------------------------------------------------------------------------
Underidentification test (Kleibergen-Paap rk LM statistic): 18.031
Chi-sq(1) P-val = 0.0000
------------------------------------------------------------------------------
Weak identification test (Cragg-Donald Wald F statistic): 228.980
(Kleibergen-Paap rk Wald F statistic): 36.294
Stock-Yogo weak ID test critical values: 10% maximal IV size 16.38
15% maximal IV size 8.96
20% maximal IV size 6.66
25% maximal IV size 5.53
Source: Stock-Yogo (2005). Reproduced by permission.
NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors.
------------------------------------------------------------------------------
Hansen J statistic (overidentification test of all instruments): 0.000
(equation exactly identified)
-endog- option:
Endogeneity test of endogenous regressors: 10.926
Chi-sq(1) P-val = 0.0009
Regressors tested: bm
------------------------------------------------------------------------------
Instrumented: bm
Included instruments: lnpn polity2 popd
Excluded instruments: msphund
Partialled-out: religion 2.colonyg 3.colonyg 4.colonyg 5.colonyg
6.colonyg 2.ccode 3.ccode 4.ccode 5.ccode 6.ccode 7.ccode
8.ccode 9.ccode 10.ccode 11.ccode 12.ccode 13.ccode
15.ccode 17.ccode 18.ccode 19.ccode 20.ccode 21.ccode
22.ccode 23.ccode 24.ccode 25.ccode 26.ccode 27.ccode
28.ccode 29.ccode 30.ccode 31.ccode 32.ccode 34.ccode
35.ccode 37.ccode 39.ccode 41.ccode 42.ccode _cons
nb: total SS, model F and R2s are after partialling-out;
any small-sample adjustments include partialled-out
variables in regressor count K
Dropped collinear: 14.ccode 16.ccode 33.ccode 36.ccode 40.ccode 43.ccode
------------------------------------------------------------------------------
.
end of do-file
.
does it mean that the coefficient will not be changed for those that are dropped in the partial out model... can i merge the results for instance the coefficients table from the first model and postestimation results from the partial out model and then provide a note "post-estimation statistics are obtained using Frisch-Waugh-Lovell Theorem to partial out the exogenous regressors"
Comment