I am trying to analyze the effect of treatment (binary 0,1/fen,Dexmed - trt) with repeated measures (time - from 0 to 60 minutes, at 10 minute intervals). The outcome variable (rass) is not normally distributed and I chose to go with GEE. ANOVA in stata provides a p-value for trt#time but GEE provides separate p-values for each category of trt interacting with each instance of time. Is there anything wrong with my code or do I need to do something further?
Code:
xtgee rass b(last).trt time b(last).trt##time, family(gaussian) link(identity) corr(exchangeable)
note: 60.time omitted because of collinearity
Iteration 1: tolerance = .02042108
Iteration 2: tolerance = .00005279
Iteration 3: tolerance = 1.401e-07
GEE population-averaged model Number of obs = 433
Group variable: id Number of groups = 72
Link: identity Obs per group:
Family: Gaussian min = 5
Correlation: exchangeable avg = 6.0
max = 7
Wald chi2(13) = 276.36
Scale parameter: .2798006 Prob > chi2 = 0.0000
------------------------------------------------------------------------------
rass | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
trt |
Dexmed | -.75 .1246775 -6.02 0.000 -.9943634 -.5056366
time | -.0019698 .0027288 -0.72 0.470 -.0073181 .0033785
|
time |
10 | -.3136353 .1104463 -2.84 0.005 -.530106 -.0971645
20 | -1.016159 .1101884 -9.22 0.000 -1.232125 -.8001941
30 | -.9686835 .1165067 -8.31 0.000 -1.197032 -.7403346
40 | -.754541 .1284343 -5.87 0.000 -1.006268 -.5028144
50 | -.3401015 .1508843 -2.25 0.024 -.6358293 -.0443737
60 | 0 (omitted)
|
trt#time |
Dexmed#10 | -.3333333 .1657982 -2.01 0.044 -.6582918 -.0083748
Dexmed#20 | .4166667 .1657982 2.51 0.012 .0917082 .7416252
Dexmed#30 | .5 .1657982 3.02 0.003 .1750415 .8249585
Dexmed#40 | .3611111 .1657982 2.18 0.029 .0361526 .6860696
Dexmed#50 | .350862 .1763269 1.99 0.047 .0052677 .6964563
Dexmed#60 | .0793486 .4073754 0.19 0.846 -.7190925 .8777898
|
_cons | .8333333 .0881603 9.45 0.000 .6605423 1.006124
------------------------------------------------------------------------------
Comment