在进行中介效应检验时,用固定效应模型(xtreg y x1 x2,fe)采用三步逐步回归,发现自变量和中介变量都是显著的,按理来说就存在中介效应了;但是用sobel和bootstrap检验时,中介变量就不显著了,根据bootstrap得出来的95%置信区间里面也包含了0,通过对比系数发现sobel和bootstrap检验时跑出来的回归和混合回归(OLS)基本一样,请问这时候要选哪个方法啊?sobel和bootstrap检验时能使用固定效应回归(fe)吗,还是它本身就是OLS回归啊?如下:因变量(xHQD)、自变量(LnAI专申)、中介变量(indh)三步回归
. xtreg xHQD LnAI专申 Gov Urban Ope InfLn Fin i.year,fe
----------------------------------------------------------------------------
xHQD | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
LnAI专申 | 1.031515 .2480856 4.16 0.000 .5426579 1.520371 (自变量对因变量的总效应显著)
Gov | -1.950249 7.078815 -0.28 0.783 -15.89917 11.99867
. xtreg indh LnAI专申 Gov Urban Ope InfLn Fin i.year,fe
------------------------------------------------------------------------------
indh | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
LnAI专申 | .7073071 .2803669 2.52 0.012 .1548396 1.259775(自变量对中介显著)
. xtreg xHQD indh LnAI专申 Gov Urban Ope InfLn Fin i.year,fe
------------------------------------------------------------------------------
xHQD | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
indh | .443379 .0510509 8.69 0.000 .34278 .543978
LnAI专申 | .7179094 .2181799 3.29 0.001 .2879721 1.147847 (中介和自变量对因变量都显著)
Sobel检验:
. sgmediation xHQD, mv(indh) iv(LnAI专申) cv(Gov Urban Ope InfLn Fin year1-year9
Model with dv regressed on iv (path c)
Source | SS df MS Number of obs = 270
-------------+---------------------------------- F(14, 255) = 114.93
Model | 20815.7103 14 1486.83645 Prob > F = 0.0000
Residual | 3298.78107 255 12.9363964 R-squared = 0.8632
-------------+---------------------------------- Adj R-squared = 0.8557
Total | 24114.4914 269 89.6449494 Root MSE = 3.5967
------------------------------------------------------------------------------
xHQD | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
LnAI专申 | 2.353209 .3329573 7.07 0.000 1.697513 3.008906 (显著)
Model with mediator regressed on iv (path a)
Source | SS df MS Number of obs = 270
-------------+---------------------------------- F(14, 255) = 18.32
Model | 5025.02978 14 358.930698 Prob > F = 0.0000
Residual | 4997.15139 255 19.5966721 R-squared = 0.5014
-------------+---------------------------------- Adj R-squared = 0.4740
Total | 10022.1812 269 37.2571791 Root MSE = 4.4268
------------------------------------------------------------------------------
indh | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
LnAI专申 | -.1517652 .4098011 -0.37 0.711 -.9587908 .6552604 (不显著了??)
Model with dv regressed on mediator and iv (paths b and c')
Source | SS df MS Number of obs = 270
-------------+---------------------------------- F(15, 254) = 115.17
Model | 21023.4376 15 1401.56251 Prob > F = 0.0000
Residual | 3091.05379 254 12.1695031 R-squared = 0.8718
-------------+---------------------------------- Adj R-squared = 0.8642
Total | 24114.4914 269 89.6449494 Root MSE = 3.4885
------------------------------------------------------------------------------
xHQD | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
indh | .2038851 .0493486 4.13 0.000 .1067005 .3010697
LnAI专申 | 2.384152 .3230242 7.38 0.000 1.748005 3.020299(都显著)
Sobel-Goodman Mediation Tests
Coef Std Err Z P>|Z|
Sobel -.03094267 .08388733 -.3689 .71223216(不显著)
Goodman-1 (Aroian) -.03094267 .08629055 -.3586 .71990405
Goodman-2 -.03094267 .0814132 -.3801 .70389385
Coef Std Err Z P>|Z|
a coefficient = -.151765 .409801 -.370339 .71113
b coefficient = .203885 .049349 4.13152 .000036
Indirect effect = -.030943 .083887 -.36886 .712232
Direct effect = 2.38415 .323024 7.38072 1.6e-13
Total effect = 2.35321 .332957 7.0676 1.6e-12
Proportion of total effect that is mediated: -.01314914
Ratio of indirect to direct effect: -.01297848
Ratio of total to direct effect: .98702152
Bootstrap results Number of obs = 270
Replications = 220
. bootstrap r(ind_eff) r(dir_eff), reps(1000): sgmediation xHQD , mv(indh) iv(LnAI专申) cv(Gov Urban Ope InfLn Fin year1-year9)
command: sgmediation xHQD, mv(indh) iv(LnAI专申) cv(Gov Urban Ope InfLn
Fin year1-year9)
_bs_1: r(ind_eff)
_bs_2: r(dir_eff)
------------------------------------------------------------------------------
| Observed Bootstrap Normal-based
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
_bs_1 | -.0309427 .0908348 -0.34 0.733 -.2089757 .1470904(包含0)
_bs_2 | 2.384152 .3546166 6.72 0.000 1.689116 3.079188
------------------------------------------------------------------------------


雷达卡





京公网安备 11010802022788号







