dear statalisters:
i have attempted to use margins with contrast operators to compare all levels of a 3+ level categorical variable to the base level after anova or xtreg. this works fine with r.factor (assuming first level is base), but does not work with r2.factor, which, according to documentation, should set the second level as the base category for comparison.
i may just be misinterpretting margins, contrast documentation.
see code below
could you please confirm my error in understanding or an error in Stata?
i have attempted to use margins with contrast operators to compare all levels of a 3+ level categorical variable to the base level after anova or xtreg. this works fine with r.factor (assuming first level is base), but does not work with r2.factor, which, according to documentation, should set the second level as the base category for comparison.
i may just be misinterpretting margins, contrast documentation.
see code below
Code:
. sysuse bplong
(fictional blood-pressure data)
. xtset patient when
panel variable: patient (strongly balanced)
time variable: when, 1 to 2
delta: 1 unit
. xtdes
patient: 1, 2, ..., 120 n = 120
when: 1, 2, ..., 2 T = 2
Delta(when) = 1 unit
Span(when) = 2 periods
(patient*when uniquely identifies each observation)
Distribution of T_i: min 5% 25% 50% 75% 95% max
2 2 2 2 2 2 2
Freq. Percent Cum. | Pattern
---------------------------+---------
120 100.00 100.00 | 11
---------------------------+---------
120 100.00 | XX
. xtreg bp i.sex#i.agegrp
Random-effects GLS regression Number of obs = 240
Group variable: patient Number of groups = 120
R-sq: Obs per group:
within = . min = 2
between = 0.4393 avg = 2.0
overall = 0.2442 max = 2
Wald chi2(5) = 75.59
corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000
-------------------------------------------------------------------------------
bp | Coef. Std. Err. z P>|z| [95% Conf. Interval]
--------------+----------------------------------------------------------------
sex#agegrp |
Male#46-59 | 8.2 2.570528 3.19 0.001 3.161858 13.23814
Male#60+ | 14.125 2.570528 5.49 0.000 9.086858 19.16314
Female#30-45 | -3.9 2.570528 -1.52 0.129 -8.938142 1.138142
Female#46-59 | -2.225 2.570528 -0.87 0.387 -7.263142 2.813142
Female#60+ | 7.525 2.570528 2.93 0.003 2.486858 12.56314
|
_cons | 149.95 1.817638 82.50 0.000 146.3875 153.5125
--------------+----------------------------------------------------------------
sigma_u | 0
sigma_e | 12.307349
rho | 0 (fraction of variance due to u_i)
-------------------------------------------------------------------------------
. margins i.agegrp
Predictive margins Number of obs = 240
Model VCE : Conventional
Expression : Linear prediction, predict()
------------------------------------------------------------------------------
| Delta-method
| Margin Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
agegrp |
30-45 | 148 1.285264 115.15 0.000 145.4809 150.5191
46-59 | 152.9375 1.285264 118.99 0.000 150.4184 155.4566
60+ | 160.775 1.285264 125.09 0.000 158.2559 163.2941
------------------------------------------------------------------------------
. margins r.agegrp
Contrasts of predictive margins
Model VCE : Conventional
Expression : Linear prediction, predict()
-----------------------------------------------------
| df chi2 P>chi2
------------------+----------------------------------
agegrp |
(46-59 vs 30-45) | 1 7.38 0.0066
(60+ vs 30-45) | 1 49.40 0.0000
Joint | 2 50.25 0.0000
-----------------------------------------------------
-------------------------------------------------------------------
| Delta-method
| Contrast Std. Err. [95% Conf. Interval]
------------------+------------------------------------------------
agegrp |
(46-59 vs 30-45) | 4.9375 1.817638 1.374996 8.500004
(60+ vs 30-45) | 12.775 1.817638 9.212496 16.3375
-------------------------------------------------------------------
. margins r2.agegrp
invalid contrast;
factor with only one level not allowed
r(322);
.
Comment