Thus i decided to run an Oaxaca-Blinder threefold decomposition, and the results are the following:
Code:
oaxaca creativity age edu city sex openness, by(media_use) noisily Model for group 1 Source | SS df MS Number of obs = 1099 -------------+------------------------------ F( 5, 1093) = 11.27 Model | 139.229392 5 27.8458784 Prob > F = 0.0000 Residual | 2701.30382 1093 2.47145821 R-squared = 0.0490 -------------+------------------------------ Adj R-squared = 0.0447 Total | 2840.53321 1098 2.58700657 Root MSE = 1.5721 ------------------------------------------------------------------------------ creativity | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- age | -.0220914 .0062677 -3.52 0.000 -.0343895 -.0097934 edu | .0171041 .0744698 0.23 0.818 -.1290158 .163224 city | .0119995 .0059058 2.03 0.042 .0004116 .0235875 sex | .2500307 .0954181 2.62 0.009 .0628074 .4372541 openness | .1069547 .0172714 6.19 0.000 .0730659 .1408435 _cons | 7.766424 .3895031 19.94 0.000 7.002166 8.530683 ------------------------------------------------------------------------------ Model for group 2 Source | SS df MS Number of obs = 268 -------------+------------------------------ F( 5, 262) = 3.70 Model | 29.3619807 5 5.87239615 Prob > F = 0.0030 Residual | 415.455183 262 1.58570681 R-squared = 0.0660 -------------+------------------------------ Adj R-squared = 0.0482 Total | 444.817164 267 1.66598189 Root MSE = 1.2592 ------------------------------------------------------------------------------ creativity | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- age | -.0269685 .0100983 -2.67 0.008 -.0468526 -.0070843 edu | .0075992 .1059643 0.07 0.943 -.2010508 .2162492 city | .0014394 .0100752 0.14 0.887 -.0183993 .021278 sex | .3840037 .1559485 2.46 0.014 .0769317 .6910757 openness | .0512226 .0284962 1.80 0.073 -.0048881 .1073332 _cons | 9.151029 .6641105 13.78 0.000 7.843356 10.4587 ------------------------------------------------------------------------------ Blinder-Oaxaca decomposition Number of obs = 1367 Model = linear Group 1: media_use = 0 N of obs 1 = 1099 Group 2: media_use = 1 N of obs 2 = 268 ------------------------------------------------------------------------------ creativity | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- overall | group_1 | 8.621474 .0486231 177.31 0.000 8.526175 8.716774 group_2 | 9.026119 .0795434 113.47 0.000 8.870217 9.182022 difference | -.4046453 .0932274 -4.34 0.000 -.5873678 -.2219229 endowments | -.0504749 .0403209 -1.25 0.211 -.1295025 .0285526 coefficients | -.3243429 .0939862 -3.45 0.001 -.5085525 -.1401334 interaction | -.0298274 .0427209 -0.70 0.485 -.1135589 .0539041 -------------+---------------------------------------------------------------- endowments | age | .0049286 .0142001 0.35 0.729 -.0229031 .0327603 edu | .0018419 .0256868 0.07 0.943 -.0485033 .0521871 city | .0015615 .0109589 0.14 0.887 -.0199176 .0230406 sex | -.0148148 .0144081 -1.03 0.304 -.0430542 .0134246 openness | -.0439922 .0263984 -1.67 0.096 -.0957321 .0077477 -------------+---------------------------------------------------------------- coefficients | age | .1840348 .4484962 0.41 0.682 -.6950016 1.063071 edu | .0199319 .2715956 0.07 0.941 -.5123857 .5522495 city | .2067904 .2287483 0.90 0.366 -.241548 .6551288 sex | -.0664866 .0908223 -0.73 0.464 -.244495 .1115219 openness | .7159911 .4281922 1.67 0.094 -.1232501 1.555232 _cons | -1.384605 .7699061 -1.80 0.072 -2.893593 .1243836 -------------+---------------------------------------------------------------- interaction | age | -.0008913 .0033468 -0.27 0.790 -.0074509 .0056683 edu | .0023038 .0313959 0.07 0.942 -.0592311 .0638387 city | .0114565 .0139335 0.82 0.411 -.0158525 .0387656 sex | .0051687 .0084031 0.62 0.538 -.0113011 .0216384 openness | -.0478652 .0305762 -1.57 0.117 -.1077934 .0120631 ------------------------------------------------------------------------------
Specifically, by looking at the estimation associated to 'endowments' and 'coefficients' (in bold), may I assume that the differences arising from media_use are entirely attributable to the way people interact with media, instead of a different distribution of endowment by media_use? Should this suggest that I do not incur in any selection bias in considering the association between my X and Y?
thanks a lot in advance for your tips.
Best, G
0 Response to Oaxaca threefold decomposition and selection: how to interpret
Post a Comment