Dear statalisters,
I am running a model on a panel with two time points and want to
estimate simultaneously fixed effects (within) and between effects
using this sintax which follows a procedure suggested in Sophia
Rabe-Hesketh´s book on multilevel and lingitudinal modelling using
stata 2nd ed.
egen mean_bpd=mean(bpd), by(id)
egen delta_bpd= bpd-mean_bpd
xtreg dep mean_bpd delta_bpd
By doing this, mean_bpd coefficient is the between effect and
delta_bpd is the fixed effect.
The output is :
. xtreg dep nm_bp dev_bp
Random-effects GLS regression Number of obs = 407
Group variable: subject Number of groups = 270
R-sq: within = 0.1137 Obs per group: min = 1
between = 0.0913 avg = 1.5
overall = 0.0906 max = 2
Random effects u_i ~ Gaussian Wald chi2(2) = 43.69
corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000
dep Coef. Std. Err. z P>z [95% Conf. Interval]
nm_bp 1.135272 .2410716 4.71 0.000 .6627809 1.607764
dev_bp .8803003 .1789422 4.92 0.000 .5295801 1.23102
_cons .5785817 .1756038 3.29 0.001 .2344046 .9227588
sigma_u 1.0789156
sigma_e 1.1515882
rho .46745328 (fraction of variance due to u_i)
I have two questions:
1) How exactly is the nm_bp interpreted? I mean, I understand that is
the between effect subject, but not sure how to interpret it. is that
the change for "dep" in which time point? I know this may be a silly
question but it is puzzling me.
2) can the same model be run with a dichotomous dependent variable,
using xtlogit? If not, is there a way to do this?
Many thanks
Filipa
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
