After running xtreg on a model with a continuous dependent variable, I want to test the differences between levels of the intervention treatment. What is the difference between running margins VS contrast?
I see the first uses a chi test and the second uses an F test. The only real difference in results is the p-value for the comparison of level 2 vs 1 in the first is .0733, while it is .08 in the second.
Which command should I use? In this case, there is little difference, but could there be a case where the p-value difference between the two is substantial for interpretation?
I see the first uses a chi test and the second uses an F test. The only real difference in results is the p-value for the comparison of level 2 vs 1 in the first is .0733, while it is .08 in the second.
Which command should I use? In this case, there is little difference, but could there be a case where the p-value difference between the two is substantial for interpretation?
Code:
. margins ar.interven123
Contrasts of predictive margins
Model VCE : Robust
Expression : Linear prediction, predict()
df chi2 P>chi2
interven123
(1 vs 0) 1 0.02 0.8963
(2 vs 1) 1 3.21 0.0733
(3 vs 2) 1 12.00 0.0005
Joint 3 17.54 0.0005
Delta-method
Contrast Std. Err. [95% Conf. Interval]
interven123
(1 vs 0) -1735.885 13315.37 -27833.52 24361.75
(2 vs 1) 37908 21165.87 -3576.35 79392.35
(3 vs 2) -21796.77 6293.4 -34131.6 -9461.928
. contrast ar.interven123, effects
Contrasts of marginal linear predictions
Margins : asbalanced
df F P>F
interven123
(1 vs 0) 1 0.02 0.8969
(2 vs 1) 1 3.21 0.0802
(3 vs 2) 1 12.00 0.0012
Joint 3 5.85 0.0019
Denominator 44
Contrast Std. Err. t P>t [95% Conf. Interval]
interven123
(1 vs 0) -1735.885 13315.37 -0.13 0.897 -28571.24 25099.47
(2 vs 1) 37908 21165.87 1.79 0.080 -4749.015 80565.01
(3 vs 2) -21796.77 6293.4 -3.46 0.001 -34480.28 -9113.251
.
end of do-file
