Dear all,
I want to use the command--sgmediation- to analyze the mediation of panel data.I read that --sgmediation--can not work for panel data.But I think that this command also works for panel data.
Here is my codes.Do you think it is right or wrong?
This is just a very simple example. The data I input at will.Do you think the codes is right or wrong?I just put individual fixed effects and time fixed effects in cv( ).But in the bootstrap,there are many "xxxxxxxxxxxx......".I don't know why.
Best
Raymond
I want to use the command--sgmediation- to analyze the mediation of panel data.I read that --sgmediation--can not work for panel data.But I think that this command also works for panel data.
Here is my codes.Do you think it is right or wrong?
Code:
* Example generated by -dataex-. To install: ssc install dataex
clear
input double(id year y x1 x2 x3)
1 1 3 2 7 3
1 2 2 1 6 4
1 3 3 2 6 5
1 4 4 2 5 6
1 5 5 3 4 7
2 1 6 4 5 8
2 2 5 3 6 8
2 3 5 3 5 7
2 4 4 2 4 6
2 5 3 3 5 5
3 1 4 5 6 4
3 2 5 6 7 3
3 3 6 7 8 6
3 4 5 8 8 7
3 5 4 6 7 8
3 6 3 4 6 9
4 1 6 5 5 8
4 2 5 6 4 7
4 3 6 7 3 6
4 4 7 7 . 5
4 5 8 6 2 4
4 6 9 5 4 3
end
xi:sgmediation y,mv(x2) iv(x1) cv(x3 i.id i.year)
i.id _Iid_1-4 (naturally coded; _Iid_1 omitted)
i.year _Iyear_1-6 (naturally coded; _Iyear_1 omitted)
Model with dv regressed on iv (path c)
Source | SS df MS Number of obs = 21
-------------+---------------------------------- F(10, 10) = 2.63
Model | 41.4558892 10 4.14558892 Prob > F = 0.0718
Residual | 15.7822061 10 1.57822061 R-squared = 0.7243
-------------+---------------------------------- Adj R-squared = 0.4485
Total | 57.2380952 20 2.86190476 Root MSE = 1.2563
------------------------------------------------------------------------------
y | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
x1 | 0.896 0.398 2.25 0.048 0.009 1.782
x3 | -0.170 0.164 -1.04 0.325 -0.534 0.195
_Iid_2 | 0.610 0.922 0.66 0.523 -1.445 2.664
_Iid_3 | -2.615 1.866 -1.40 0.191 -6.773 1.542
_Iid_4 | -0.291 1.857 -0.16 0.879 -4.429 3.846
_Iyear_2 | -0.542 0.889 -0.61 0.556 -2.524 1.439
_Iyear_3 | -0.379 0.937 -0.41 0.694 -2.466 1.707
_Iyear_4 | -0.223 1.005 -0.22 0.829 -2.462 2.016
_Iyear_5 | -0.155 0.910 -0.17 0.868 -2.184 1.873
_Iyear_6 | 1.724 1.254 1.37 0.199 -1.071 4.518
_cons | 2.717 1.254 2.17 0.056 -0.077 5.510
------------------------------------------------------------------------------
Model with mediator regressed on iv (path a)
Source | SS df MS Number of obs = 21
-------------+---------------------------------- F(10, 10) = 3.55
Model | 38.1930886 10 3.81930886 Prob > F = 0.0290
Residual | 10.7592924 10 1.07592924 R-squared = 0.7802
-------------+---------------------------------- Adj R-squared = 0.5604
Total | 48.952381 20 2.44761905 Root MSE = 1.0373
------------------------------------------------------------------------------
x2 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
x1 | 0.307 0.328 0.94 0.372 -0.425 1.039
x3 | 0.014 0.135 0.11 0.918 -0.287 0.315
_Iid_2 | -0.933 0.761 -1.23 0.248 -2.629 0.763
_Iid_3 | 0.122 1.541 0.08 0.938 -3.310 3.555
_Iid_4 | -3.274 1.533 -2.14 0.059 -6.690 0.143
_Iyear_2 | 0.004 0.734 0.00 0.996 -1.632 1.640
_Iyear_3 | -0.484 0.773 -0.63 0.545 -2.207 1.239
_Iyear_4 | -0.843 0.830 -1.02 0.334 -2.691 1.006
_Iyear_5 | -1.407 0.752 -1.87 0.091 -3.082 0.267
_Iyear_6 | -0.353 1.036 -0.34 0.741 -2.660 1.955
_cons | 5.460 1.035 5.27 0.000 3.153 7.767
------------------------------------------------------------------------------
Model with dv regressed on mediator and iv (paths b and c')
Source | SS df MS Number of obs = 21
-------------+---------------------------------- F(11, 9) = 2.24
Model | 41.9107839 11 3.81007126 Prob > F = 0.1184
Residual | 15.3273113 9 1.70303459 R-squared = 0.7322
-------------+---------------------------------- Adj R-squared = 0.4049
Total | 57.2380952 20 2.86190476 Root MSE = 1.305
------------------------------------------------------------------------------
y | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
x2 | -0.206 0.398 -0.52 0.618 -1.106 0.694
x1 | 0.959 0.431 2.23 0.053 -0.016 1.934
x3 | -0.167 0.170 -0.98 0.353 -0.551 0.218
_Iid_2 | 0.418 1.027 0.41 0.694 -1.906 2.741
_Iid_3 | -2.590 1.939 -1.34 0.214 -6.976 1.796
_Iid_4 | -0.964 2.328 -0.41 0.688 -6.229 4.301
_Iyear_2 | -0.542 0.924 -0.59 0.572 -2.631 1.548
_Iyear_3 | -0.479 0.992 -0.48 0.641 -2.722 1.765
_Iyear_4 | -0.397 1.096 -0.36 0.726 -2.877 2.084
_Iyear_5 | -0.445 1.099 -0.40 0.695 -2.931 2.041
_Iyear_6 | 1.651 1.311 1.26 0.239 -1.313 4.616
_cons | 3.839 2.533 1.52 0.164 -1.891 9.569
------------------------------------------------------------------------------
Sobel-Goodman Mediation Tests
Coef Std Err Z P>|Z|
Sobel -.0631524 .13961254 -.4523 .65102373
Goodman-1 (Aroian) -.0631524 .19122088 -.3303 .74120433
Goodman-2 -.0631524 .04917211 -1.284 .19903234
Coef Std Err Z P>|Z|
a coefficient = .307133 .328431 .93515 .349711
b coefficient = -.205619 .39785 -.516825 .605278
Indirect effect = -.063152 .139613 -.45234 .651024
Direct effect = .958806 .430893 2.22516 .02607
Total effect = .895654 .397774 2.25166 .024344
Proportion of total effect that is mediated: -.07050982
Ratio of indirect to direct effect: -.06586565
Ratio of total to direct effect: .93413435
tab id,gen(dum_id)
id | Freq. Percent Cum.
------------+-----------------------------------
1 | 5 22.73 22.73
2 | 5 22.73 45.45
3 | 6 27.27 72.73
4 | 6 27.27 100.00
------------+-----------------------------------
Total | 22 100.00
. tab year,gen(dum_year)
year | Freq. Percent Cum.
------------+-----------------------------------
1 | 4 18.18 18.18
2 | 4 18.18 36.36
3 | 4 18.18 54.55
4 | 4 18.18 72.73
5 | 4 18.18 90.91
6 | 2 9.09 100.00
------------+-----------------------------------
Total | 22 100.00
. bootstrap r(ind_eff) r(dir_eff), reps(500) : sgmediation y, mv(x2) iv(x1) cv(x3 dum_id* dum_year*)
(running sgmediation on estimation sample)
Bootstrap replications (500)
----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5
xxxxxxxxxxxxxxxxxx.xxxxxxxxxxxxxxxxxxxxxxxxxx.x.xx 50
xxxxx.xxxxxxxxxxxxxxxxxxx.xxxxxxxxxxxxxxxxxxxxxxxx 100
xxxxxxxxx.xxxxxxxxxx.xx.xxxxxxxxx.xxx.xx.xxxxxx.xx 150
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 200
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx.xxxxxxxxx. 250
xxxxxx.xxxxxxx.xx..xxxxxxxx.xxxxxxxxxx..xxxxxxxxxx 300
xxxxxxxxxxxxxxxxx.xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 350
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 400
.xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx.xxxxx 450
xxxxxxxxxx.xxxxxxxxxxxxxxxxxxx.xxxxxxxxxxxx.x.x..x 500
Bootstrap results Number of obs = 21
Replications = 30
command: sgmediation y, mv(x2) iv(x1) cv(x3 dum_id* dum_year*)
_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 | -0.063 0.641 -0.10 0.921 -1.319 1.192
_bs_2 | 0.959 0.911 1.05 0.293 -0.828 2.745
------------------------------------------------------------------------------
Note: One or more parameters could not be estimated in 470 bootstrap
replicates; standard-error estimates include only complete replications.
Best
Raymond
Comment