sorry forgot to mention, that the variable names this is concerning is epad (automotive diesel) and epop (oil products)
-
Login or Register
- Log in with
- You are not logged in. You can browse but not post. Login or Register by clicking 'Login or Register' at the top-right of this page. For more information on Statalist, see the FAQ.
-
-
Anne:
some comments about your query:
1) it may be due to heteroskedasticity (hence, cluster-robust standard errors are the way to go, even if your groups are only 20);
2) it may be due to too many predictors in the right-hand side of your regression equation (some of them seem redundant).Kind regards,
Carlo
(StataNow 18.5)Comment
-
Carlo, thank you! I had completely overlooked, that I didn't get cluster-robust std. errors.
Would you recommend me performing a wald test on the predictor variables, to figure out if some of my variables are redundant? Is this a compatible method with re?Code:. xtreg bincw povri dipa ddpa noren noown hhgp hhep epad epop epug wgge wgphs wgsst rec, re cluster() Random-effects GLS regression Number of obs = 209 Group variable: cntry Number of groups = 20 R-squared: Obs per group: Within = 0.5041 min = 6 Between = 0.7673 avg = 10.4 Overall = 0.7141 max = 11 Wald chi2(14) = 295.04 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ bincw | Coefficient Std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- povri | 1.077634 .10159 10.61 0.000 .8785216 1.276747 dipa | -.0344149 .0358914 -0.96 0.338 -.1047607 .0359309 ddpa | .0040692 .0350314 0.12 0.908 -.064591 .0727294 noren | -.443122 2.643103 -0.17 0.867 -5.623508 4.737264 noown | -7.269417 2.877654 -2.53 0.012 -12.90952 -1.629318 hhgp | .2621671 .2772514 0.95 0.344 -.2812356 .8055698 hhep | .0435926 .1259867 0.35 0.729 -.2033368 .2905219 epad | -.2321981 .0585484 -3.97 0.000 -.3469508 -.1174453 epop | -.1155069 .1115238 -1.04 0.300 -.3340895 .1030757 epug | .4527002 .1237078 3.66 0.000 .2102374 .695163 wgge | .017569 .0935747 0.19 0.851 -.1658341 .2009721 wgphs | -.0157227 .095093 -0.17 0.869 -.2021015 .1706561 wgsst | -.2055807 .2426368 -0.85 0.397 -.6811401 .2699787 rec | -.125298 .054656 -2.29 0.022 -.2324219 -.0181742 _cons | 3.483138 8.441969 0.41 0.680 -13.06282 20.02909 -------------+---------------------------------------------------------------- sigma_u | 2.7264747 sigma_e | 2.3243363 rho | .57911706 (fraction of variance due to u_i) ------------------------------------------------------------------------------ . xtreg bincw povri dipa ddpa noren noown hhgp hhep epad epop epug wgge wgphs wgsst rec, re vce(cluster cntry) Random-effects GLS regression Number of obs = 209 Group variable: cntry Number of groups = 20 R-squared: Obs per group: Within = 0.5041 min = 6 Between = 0.7673 avg = 10.4 Overall = 0.7141 max = 11 Wald chi2(14) = 741.71 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 (Std. err. adjusted for 20 clusters in cntry) ------------------------------------------------------------------------------ | Robust bincw | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- povri | 1.077634 .1645577 6.55 0.000 .7551071 1.400162 dipa | -.0344149 .0399869 -0.86 0.389 -.1127877 .043958 ddpa | .0040692 .0303355 0.13 0.893 -.0553873 .0635257 noren | -.443122 2.420777 -0.18 0.855 -5.187757 4.301513 noown | -7.269417 2.72881 -2.66 0.008 -12.61779 -1.921048 hhgp | .2621671 .3862824 0.68 0.497 -.4949324 1.019267 hhep | .0435926 .1389173 0.31 0.754 -.2286803 .3158655 epad | -.2321981 .0671894 -3.46 0.001 -.3638869 -.1005092 epop | -.1155069 .0935965 -1.23 0.217 -.2989526 .0679388 epug | .4527002 .121722 3.72 0.000 .2141295 .6912709 wgge | .017569 .0910043 0.19 0.847 -.1607962 .1959342 wgphs | -.0157227 .1796008 -0.09 0.930 -.3677338 .3362885 wgsst | -.2055807 .3137024 -0.66 0.512 -.8204261 .4092648 rec | -.125298 .0752764 -1.66 0.096 -.2728371 .0222411 _cons | 3.483138 16.28392 0.21 0.831 -28.43276 35.39903 -------------+---------------------------------------------------------------- sigma_u | 2.7264747 sigma_e | 2.3243363 rho | .57911706 (fraction of variance due to u_i) ------------------------------------------------------------------------------
BR AnneComment
-
Anne:
as some p-values are alarmingly high, I would recommend you to check the correlation of the coefficients via:
Code:estat vce, corr
Kind regards,
Carlo
(StataNow 18.5)Comment
-
I think i attached the data with a new variable (povri) that shouldn't be included. so the true results is as follows:
I tried running the wald test on my control variables and got the following results.Code:xtreg bincw dipa ddpa noren noown hhgp hhep epad epop epug wgge wgphs wgsst rec, re vce(cluster cntry) Random-effects GLS regression Number of obs = 209 Group variable: cntry Number of groups = 20 R-squared: Obs per group: Within = 0.1957 min = 6 Between = 0.7145 avg = 10.4 Overall = 0.6331 max = 11 Wald chi2(13) = 164.14 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 (Std. err. adjusted for 20 clusters in cntry) ------------------------------------------------------------------------------ | Robust bincw | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- dipa | -.11896 .070824 -1.68 0.093 -.2577725 .0198524 ddpa | .0269258 .0503971 0.53 0.593 -.0718507 .1257023 noren | -1.45799 3.079142 -0.47 0.636 -7.492997 4.577017 noown | -8.542249 3.339083 -2.56 0.011 -15.08673 -1.997767 hhgp | .3165467 .4411764 0.72 0.473 -.5481432 1.181237 hhep | -.0073165 .2191188 -0.03 0.973 -.4367815 .4221485 epad | -.2064444 .0897387 -2.30 0.021 -.3823291 -.0305598 epop | -.2397357 .1423787 -1.68 0.092 -.5187928 .0393214 epug | .611444 .1795668 3.41 0.001 .2594995 .9633884 wgge | .1650103 .1214228 1.36 0.174 -.072974 .4029946 wgphs | -.4320382 .188689 -2.29 0.022 -.8018619 -.0622145 wgsst | -.1977692 .3755178 -0.53 0.598 -.9337706 .5382321 rec | -.1316739 .0690174 -1.91 0.056 -.2669455 .0035976 _cons | 51.801 15.00813 3.45 0.001 22.38561 81.21639 -------------+---------------------------------------------------------------- sigma_u | 4.1444811 sigma_e | 3.0344652 rho | .6510107 (fraction of variance due to u_i) ------------------------------------------------------------------------------ . xtreg bincw dipa ddpa noren noown hhgp hhep epad epop epug wgge wgphs wgsst rec, re cluster() Random-effects GLS regression Number of obs = 209 Group variable: cntry Number of groups = 20 R-squared: Obs per group: Within = 0.1957 min = 6 Between = 0.7145 avg = 10.4 Overall = 0.6331 max = 11 Wald chi2(13) = 100.08 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ bincw | Coefficient Std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- dipa | -.11896 .0437347 -2.72 0.007 -.2046785 -.0332416 ddpa | .0269258 .0433534 0.62 0.535 -.0580454 .111897 noren | -1.45799 3.400739 -0.43 0.668 -8.123315 5.207335 noown | -8.542249 3.724434 -2.29 0.022 -15.84201 -1.242492 hhgp | .3165467 .3465723 0.91 0.361 -.3627226 .995816 hhep | -.0073165 .1588096 -0.05 0.963 -.3185776 .3039445 epad | -.2064444 .0724658 -2.85 0.004 -.3484749 -.064414 epop | -.2397357 .136564 -1.76 0.079 -.5073962 .0279248 epug | .611444 .1521178 4.02 0.000 .3132985 .9095894 wgge | .1650103 .11456 1.44 0.150 -.0595232 .3895438 wgphs | -.4320382 .108138 -4.00 0.000 -.6439849 -.2200916 wgsst | -.1977692 .3066784 -0.64 0.519 -.7988478 .4033094 rec | -.1316739 .0720012 -1.83 0.067 -.2727936 .0094458 _cons | 51.801 8.780813 5.90 0.000 34.59092 69.01108 -------------+---------------------------------------------------------------- sigma_u | 4.1444811 sigma_e | 3.0344652 rho | .6510107 (fraction of variance due to u_i) ------------------------------------------------------------------------------
and then i ran a regression without the variables which were rejected by the testCode:quietly xtreg bincw dipa ddpa noren noown hhgp hhep epad epop epug wgge wgphs wgsst rec, re vce(cluster cntry) . test epug ( 1) epug = 0 chi2( 1) = 11.59 Prob > chi2 = 0.0007 . test epad ( 1) epad = 0 chi2( 1) = 5.29 Prob > chi2 = 0.0214 . test epop ( 1) epop = 0 chi2( 1) = 2.84 Prob > chi2 = 0.0922 . test hhep ( 1) hhep = 0 chi2( 1) = 0.00 Prob > chi2 = 0.9734 . test hhgp ( 1) hhgp = 0 chi2( 1) = 0.51 Prob > chi2 = 0.4731 . test noren ( 1) noren = 0 chi2( 1) = 0.22 Prob > chi2 = 0.6359 . test noown ( 1) noown = 0 chi2( 1) = 6.54 Prob > chi2 = 0.0105 . test dipa ( 1) dipa = 0 chi2( 1) = 2.82 Prob > chi2 = 0.0930 . test ddpa ( 1) ddpa = 0 chi2( 1) = 0.29 Prob > chi2 = 0.5932
But it still seems weird with the automotive diesel price, maybe the wald test isn't that compatible with re?Code:. xtreg bincw noown epad epug wgge wgphs wgsst rec, re vce(cluster cntry) Random-effects GLS regression Number of obs = 231 Group variable: cntry Number of groups = 21 R-squared: Obs per group: Within = 0.2132 min = 11 Between = 0.7439 avg = 11.0 Overall = 0.6784 max = 11 Wald chi2(7) = 97.00 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 (Std. err. adjusted for 21 clusters in cntry) ------------------------------------------------------------------------------ | Robust bincw | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- noown | -9.780402 1.765615 -5.54 0.000 -13.24094 -6.31986 epad | -.240446 .0632929 -3.80 0.000 -.3644978 -.1163943 epug | .3978157 .0599291 6.64 0.000 .2803568 .5152746 wgge | .1708338 .1125525 1.52 0.129 -.0497651 .3914326 wgphs | -.5015039 .1865653 -2.69 0.007 -.8671652 -.1358426 wgsst | -.290375 .2950817 -0.98 0.325 -.8687244 .2879745 rec | -.0852776 .063578 -1.34 0.180 -.2098882 .039333 _cons | 57.30752 15.07934 3.80 0.000 27.75257 86.86248 -------------+---------------------------------------------------------------- sigma_u | 4.1072983 sigma_e | 3.1159158 rho | .63471188 (fraction of variance due to u_i) ------------------------------------------------------------------------------
BR AnneComment
-
Anne:
I would still recommend:
after -xtreg, re-.Code:estat vce, corr
Kind regards,
Carlo
(StataNow 18.5)Comment
-
if i understand a matrix like this correctly, then its best if the values are below 0.3 right? So my main concern should be epop which has a high value with electricity price(hhep) and unleaded gasoline(epug) as well as the relationship between wgsst and wgge. Am i understanding it correctly? this is based on the first correlation matrix depictedCode:. quietly xtreg bincw dipa ddpa noren noown hhgp hhep epad epop epug wgge wgphs wgsst rec, re vce(cluster cntry) . estat vce, correlation Correlation matrix of coefficients of xtreg model e(V) | dipa ddpa noren noown hhgp hhep epad epop epug wgge wgphs wgsst rec _cons -------------+-------------------------------------------------------------------------------------------------------------------------------------------- dipa | 1.0000 ddpa | 0.1914 1.0000 noren | -0.2742 -0.2549 1.0000 noown | -0.0898 -0.0810 -0.7883 1.0000 hhgp | -0.5144 -0.1604 0.2086 0.0330 1.0000 hhep | 0.4092 0.0340 0.1889 -0.4310 -0.4325 1.0000 epad | 0.0703 -0.1226 -0.2697 0.4269 -0.3018 -0.0822 1.0000 epop | 0.2765 -0.1567 -0.0039 -0.0479 -0.1742 0.7017 -0.2785 1.0000 epug | -0.2474 0.2431 0.0988 -0.1811 0.1987 -0.6428 -0.2074 -0.8600 1.0000 wgge | -0.0558 -0.0777 0.2690 -0.3212 0.1674 0.4044 -0.3154 0.2655 -0.1130 1.0000 wgphs | -0.3942 -0.0875 -0.1615 0.2365 0.3764 -0.1515 -0.1846 0.0306 0.1044 0.2155 1.0000 wgsst | 0.1600 0.2074 -0.2336 0.2188 -0.2621 -0.3663 0.2730 -0.2681 0.1506 -0.9172 -0.3305 1.0000 rec | 0.2264 -0.3334 -0.2148 0.1738 0.0568 -0.2142 -0.1832 0.1100 -0.0044 0.1518 -0.0059 -0.2286 1.0000 _cons | 0.2615 -0.0413 0.2253 -0.2564 -0.2556 0.1961 0.1154 0.0736 -0.2063 -0.2052 -0.9539 0.2406 0.0413 1.0000 . quietly xtreg bincw dipa ddpa noren noown hhgp hhep epad epop epug wgge wgphs wgsst rec, re . estat vce, correlation Correlation matrix of coefficients of xtreg model e(V) | dipa ddpa noren noown hhgp hhep epad epop epug wgge wgphs wgsst rec _cons -------------+-------------------------------------------------------------------------------------------------------------------------------------------- dipa | 1.0000 ddpa | 0.2220 1.0000 noren | 0.0248 -0.1601 1.0000 noown | -0.1406 0.0760 -0.7590 1.0000 hhgp | -0.0658 -0.1347 0.0254 -0.0011 1.0000 hhep | -0.0025 0.0146 -0.0239 -0.1927 -0.3595 1.0000 epad | -0.0577 0.1352 -0.0889 0.2579 -0.0265 -0.0094 1.0000 epop | 0.1273 -0.0130 -0.0236 -0.0959 -0.1853 0.2209 -0.2687 1.0000 epug | -0.0927 -0.0440 0.0699 -0.0704 0.1289 -0.2212 -0.3079 -0.8177 1.0000 wgge | 0.0416 -0.0464 -0.1982 0.1930 0.0777 0.2829 0.0538 -0.0857 0.0169 1.0000 wgphs | -0.0004 -0.0141 -0.0301 -0.1735 -0.0542 0.1027 -0.1609 0.0600 0.0703 -0.1137 1.0000 wgsst | -0.0397 0.1767 0.0788 -0.0744 -0.1764 -0.3493 0.0484 0.0193 -0.0169 -0.7933 -0.0174 1.0000 rec | 0.1363 0.1546 0.1531 -0.1070 -0.0871 -0.1273 0.1570 0.1369 -0.2398 0.0093 0.0659 -0.0108 1.0000 _cons | -0.0871 -0.2636 0.1057 -0.0771 0.1003 -0.0718 0.0364 0.0796 -0.1504 -0.1013 -0.8055 -0.0090 -0.1328 1.0000Comment
-
Anne:
a rho=|0.3| threshold value is probably too severe, but what is above rho=|0.6| should be considered with a bit of caution.Kind regards,
Carlo
(StataNow 18.5)Comment
-
Ah okay, that's nice.
Can I do anything about the variables that score above 0.6 to improve my data or should I just be attentive to it while analysing my results? I am thinking if I should remove some of the variables or just let it all stay?Comment
-
I might have found a feasible solution. If I use the total count for energy prices (eptot) and give up the nuances and use a total variable for noren/noown (notot), then i get the following results
I still have a few above 0.6, but this might be a better solution when thinking of my data size?Code:. xtreg tincw povri dipa ddpa notot eptot wgswe wgge wgphs wgsst rec, re vce(cluster cntry) Random-effects GLS regression Number of obs = 237 Group variable: cntry Number of groups = 22 R-squared: Obs per group: Within = 0.6273 min = 6 Between = 0.7228 avg = 10.8 Overall = 0.7025 max = 11 Wald chi2(10) = 380.55 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 (Std. err. adjusted for 22 clusters in cntry) ------------------------------------------------------------------------------ | Robust tincw | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- povri | .8328126 .1228118 6.78 0.000 .5921059 1.073519 dipa | -.0080761 .0300793 -0.27 0.788 -.0670305 .0508782 ddpa | .0323115 .0211811 1.53 0.127 -.0092028 .0738257 notot | -3.328045 1.486655 -2.24 0.025 -6.241835 -.4142544 eptot | .0226989 .0203902 1.11 0.266 -.0172651 .062663 wgswe | .1324678 .2119847 0.62 0.532 -.2830146 .5479501 wgge | -.0179962 .0384817 -0.47 0.640 -.0934189 .0574265 wgphs | -.0192147 .0845137 -0.23 0.820 -.1848585 .1464291 wgsst | -.2201606 .3028445 -0.73 0.467 -.8137249 .3734037 rec | .0216543 .041773 0.52 0.604 -.0602193 .1035278 _cons | -3.272447 7.710654 -0.42 0.671 -18.38505 11.84016 -------------+---------------------------------------------------------------- sigma_u | 3.3348653 sigma_e | 1.3819794 rho | .85343882 (fraction of variance due to u_i) ------------------------------------------------------------------------------ . estat vce, corr Correlation matrix of coefficients of xtreg model e(V) | povri dipa ddpa notot eptot wgswe wgge wgphs wgsst rec _cons -------------+-------------------------------------------------------------------------------------------------------------- povri | 1.0000 dipa | 0.4565 1.0000 ddpa | 0.2919 0.5822 1.0000 notot | 0.0655 -0.3291 -0.1500 1.0000 eptot | -0.6048 0.1182 0.1440 -0.2114 1.0000 wgswe | 0.0743 0.0073 -0.4271 0.0150 -0.3714 1.0000 wgge | -0.5370 -0.0253 -0.0077 -0.2415 0.1710 0.3337 1.0000 wgphs | -0.1374 -0.1301 0.3546 0.3251 -0.0478 -0.0784 0.1395 1.0000 wgsst | 0.1197 -0.0251 0.2643 -0.0067 0.2511 -0.8992 -0.6114 -0.0531 1.0000 rec | 0.3589 0.6488 0.6014 -0.2658 -0.1384 0.0048 0.0055 0.2109 -0.1457 1.0000 _cons | -0.2097 -0.1693 -0.5171 -0.4564 0.0815 0.1108 0.1167 -0.8772 -0.0785 -0.3176 1.0000
BR AnneComment
-
Anne:
your solution sounds good.
What I'd do next is investigating the correctness of the functional form of the regressand via a procedure that is explained in the -linktest- entry, Stata .manual but cannot be invoked as a built-in command after -xtreg-:
As -sq_fited- reaches statistical significance, the model is misspecified.Code:. use "https://www.stata-press.com/data/r17/nlswork.dta" (National Longitudinal Survey of Young Women, 14-24 years old in 1968) . xtreg ln_wage c.age##c.age c.tenure##c.tenure, re vce(cluster idcode) Random-effects GLS regression Number of obs = 28,101 Group variable: idcode Number of groups = 4,699 R-squared: Obs per group: Within = 0.1437 min = 1 Between = 0.2289 avg = 6.0 Overall = 0.1681 max = 15 Wald chi2(4) = 2426.91 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 (Std. err. adjusted for 4,699 clusters in idcode) ----------------------------------------------------------------------------------- | Robust ln_wage | Coefficient std. err. z P>|z| [95% conf. interval] ------------------+---------------------------------------------------------------- age | .0461889 .0041115 11.23 0.000 .0381306 .0542473 | c.age#c.age | -.0005838 .0000696 -8.38 0.000 -.0007203 -.0004473 | tenure | .0474832 .0020656 22.99 0.000 .0434347 .0515317 | c.tenure#c.tenure | -.0015861 .0001342 -11.82 0.000 -.0018491 -.0013231 | _cons | .7415106 .0582342 12.73 0.000 .6273737 .8556474 ------------------+---------------------------------------------------------------- sigma_u | .33095116 sigma_e | .29558703 rho | .55626445 (fraction of variance due to u_i) ----------------------------------------------------------------------------------- . predict fitted, xb (433 missing values generated) . g sq_fitted=fitted^2 (433 missing values generated) . xtreg ln_wage c.age##c.age c.tenure##c.tenure fitted sq_fitted , re vce(cluster idcode) note: c.tenure#c.tenure omitted because of collinearity. Random-effects GLS regression Number of obs = 28,101 Group variable: idcode Number of groups = 4,699 R-squared: Obs per group: Within = 0.1464 min = 1 Between = 0.2328 avg = 6.0 Overall = 0.1735 max = 15 Wald chi2(5) = 2538.71 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 (Std. err. adjusted for 4,699 clusters in idcode) ----------------------------------------------------------------------------------- | Robust ln_wage | Coefficient std. err. z P>|z| [95% conf. interval] ------------------+---------------------------------------------------------------- age | -.0145133 .0066019 -2.20 0.028 -.0274528 -.0015739 | c.age#c.age | .0002411 .0001001 2.41 0.016 .0000448 .0004374 | tenure | .0123218 .0029328 4.20 0.000 .0065736 .0180701 | c.tenure#c.tenure | 0 (omitted) | fitted | 5.277864 .484001 10.90 0.000 4.329239 6.226488 sq_fitted | -1.323279 .1496981 -8.84 0.000 -1.616682 -1.029876 _cons | -3.250836 .3637134 -8.94 0.000 -3.963701 -2.537971 ------------------+---------------------------------------------------------------- sigma_u | .32974949 sigma_e | .29511262 rho | .55526142 (fraction of variance due to u_i) ----------------------------------------------------------------------------------- .Kind regards,
Carlo
(StataNow 18.5)Comment
-
Is it correct if i understand this as something similar to the ramsay reset test?
Sq_fitted turned out to be insignificant in my case, but not by much.
Is this like a one-time test or should i include it in the regression output - i see that it changes some of the p values.
Code:xtreg tincw povri dipa ddpa notot eptot rec wgswe wgge wgsst wgphs, re vce(cluster cntry) Random-effects GLS regression Number of obs = 237 Group variable: cntry Number of groups = 22 R-squared: Obs per group: Within = 0.6273 min = 6 Between = 0.7228 avg = 10.8 Overall = 0.7025 max = 11 Wald chi2(10) = 380.55 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 (Std. err. adjusted for 22 clusters in cntry) ------------------------------------------------------------------------------ | Robust tincw | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- povri | .8328126 .1228118 6.78 0.000 .5921059 1.073519 dipa | -.0080761 .0300793 -0.27 0.788 -.0670305 .0508782 ddpa | .0323115 .0211811 1.53 0.127 -.0092028 .0738257 notot | -3.328045 1.486655 -2.24 0.025 -6.241835 -.4142544 eptot | .0226989 .0203902 1.11 0.266 -.0172651 .062663 rec | .0216543 .041773 0.52 0.604 -.0602193 .1035278 wgswe | .1324678 .2119847 0.62 0.532 -.2830146 .5479501 wgge | -.0179962 .0384817 -0.47 0.640 -.0934189 .0574265 wgsst | -.2201606 .3028445 -0.73 0.467 -.8137249 .3734037 wgphs | -.0192147 .0845137 -0.23 0.820 -.1848585 .1464291 _cons | -3.272447 7.710654 -0.42 0.671 -18.38505 11.84016 -------------+---------------------------------------------------------------- sigma_u | 3.3348653 sigma_e | 1.3819794 rho | .85343882 (fraction of variance due to u_i) ------------------------------------------------------------------------------ . predict fitted, xb (5 missing values generated) . g sq_fitted=fitted^2 (5 missing values generated) . xtreg tincw povri dipa ddpa notot eptot rec wgswe wgge wgsst wgphs fitted sq_fitted, re vce(cluster cntry) note: fitted omitted because of collinearity. Random-effects GLS regression Number of obs = 237 Group variable: cntry Number of groups = 22 R-squared: Obs per group: Within = 0.6584 min = 6 Between = 0.7135 avg = 10.8 Overall = 0.6985 max = 11 Wald chi2(11) = 290.51 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 (Std. err. adjusted for 22 clusters in cntry) ------------------------------------------------------------------------------ | Robust tincw | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- povri | .2368448 .2841926 0.83 0.405 -.3201625 .7938521 dipa | -.0050196 .0247749 -0.20 0.839 -.0535775 .0435383 ddpa | .0138938 .0209791 0.66 0.508 -.0272245 .0550122 notot | -1.604156 1.412062 -1.14 0.256 -4.371746 1.163434 eptot | .0139983 .0245837 0.57 0.569 -.0341849 .0621815 rec | .0191002 .0346225 0.55 0.581 -.0487586 .0869589 wgswe | .0655763 .198184 0.33 0.741 -.3228572 .4540098 wgge | .0172704 .0337686 0.51 0.609 -.0489148 .0834556 wgsst | -.1718185 .2905914 -0.59 0.554 -.7413671 .3977301 wgphs | -.0521056 .0753834 -0.69 0.489 -.1998544 .0956431 fitted | 0 (omitted) sq_fitted | .0216232 .0125354 1.72 0.085 -.0029457 .046192 _cons | 7.19012 9.232625 0.78 0.436 -10.90549 25.28573 -------------+---------------------------------------------------------------- sigma_u | 3.4893001 sigma_e | 1.3255965 rho | .87387648 (fraction of variance due to u_i) ------------------------------------------------------------------------------Comment
-
Anne:
1) do not mumble that much on sq_fitted: it does not reach ststistocal significance. Therefore, you shoud be happy, after all this going back and forth from tables, that your model is not misspecified;
2) yes, you're right it is similat to -estat ovtest- (that you cannot call after -xtreg-, though). Replicatiing your terminology, it is a one-shot test that should not be reported in the main regression otcome table. That said, I would mention in your dissertation/research report/whatever that you checked your regression for the possible misspecification of the functional form of the regressand (and that the null of no evidence of misspecification was not rejected)Kind regards,
Carlo
(StataNow 18.5)Comment
Comment