Hello all,
I'm trying to run a linear probability model with fixed effects, but I'm encountering some issues when I then try to view the predicted probabilities using 'margins'. As I understand it, margins should be based on the previous regression, and it should thus automatically include all the covariates used in that previous regression. When I run the FE regression with all covariates, and then run 'margins' afterwards, however, the predicted probabilities I am given correspond to the predicted probabilities of not having included any covariates.
I don't understand, why the predicted probabilities from margins don't fit the output I get from xtreg. I'm crossing my fingers that someone can enlighten me.
Thank you!
I'm trying to run a linear probability model with fixed effects, but I'm encountering some issues when I then try to view the predicted probabilities using 'margins'. As I understand it, margins should be based on the previous regression, and it should thus automatically include all the covariates used in that previous regression. When I run the FE regression with all covariates, and then run 'margins' afterwards, however, the predicted probabilities I am given correspond to the predicted probabilities of not having included any covariates.
I don't understand, why the predicted probabilities from margins don't fit the output I get from xtreg. I'm crossing my fingers that someone can enlighten me.
Thank you!
Code:
. xtreg purge_bi l1.i.gsdb_onset l1.(tpop_log v2x_polyarchy oil_agg latent_gdppc_mean_log), fe
> cluster(country)
Fixed-effects (within) regression Number of obs = 6,238
Group variable: country Number of groups = 165
R-sq: Obs per group:
within = 0.0171 min = 4
between = 0.3158 avg = 37.8
overall = 0.0668 max = 44
F(5,164) = 10.34
corr(u_i, Xb) = -0.1594 Prob > F = 0.0000
(Std. Err. adjusted for 165 clusters in country)
---------------------------------------------------------------------------------------
| Robust
purge_bi | Coef. Std. Err. t P>|t| [95% Conf. Interval]
----------------------+----------------------------------------------------------------
L.gsdb_onset |
1 | .0100682 .0303597 0.33 0.741 -.0498781 .0700145
|
tpop_log |
L1. | .0447499 .0212925 2.10 0.037 .0027071 .0867927
|
v2x_polyarchy |
L1. | -.309333 .0463764 -6.67 0.000 -.4009047 -.2177613
|
oil_agg |
L1. | .0002319 .001239 0.19 0.852 -.0022146 .0026783
|
latent_gdppc_mean_log |
L1. | .0032022 .0160167 0.20 0.842 -.0284232 .0348276
|
_cons | -.155505 .1794069 -0.87 0.387 -.5097502 .1987402
----------------------+----------------------------------------------------------------
sigma_u | .11790411
sigma_e | .31110025
rho | .12559436 (fraction of variance due to u_i)
---------------------------------------------------------------------------------------
. margins l1.gsdb_onset
Predictive margins Number of obs = 6,238
Model VCE : Robust
Expression : Linear prediction, predict()
------------------------------------------------------------------------------
| Delta-method
| Margin Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
L.gsdb_onset |
0 | .1323669 .0011097 119.29 0.000 .130192 .1345417
1 | .1424351 .02925 4.87 0.000 .085106 .1997641
------------------------------------------------------------------------------
Comment