Interpretation of xtgee output

[Julie Xiu]

I am trying to learn the application of xtgee to see if it is of any use in my research, and I am using Agresti Categorical Data Analysis 3rd Edition as my reference. Using the data from his Table 12.5 I get the following results:

. xtgee depression i.diagnose i.time##i.drug, family(binomial) link(logit) corr(exc) nolog

GEE population-averaged model                        Number of obs    =  1,020
Group variable: id                                   Number of groups =    340
Family: Binomial                                     Obs per group:  
Link:   Logit                                                     min =      3
Correlation: exchangeable                                         avg =    3.0
                                                                  max =      3
                                                     Wald chi2(6)     = 176.31
Scale parameter = 1                                  Prob > chi2      = 0.0000

------------------------------------------------------------------------------
  depression | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
    diagnose |
       mild  |      0.000  (base)
     severe  |     -1.312      0.146    -9.00   0.000       -1.597      -1.026
             |
        time |
          0  |      0.000  (base)
          1  |      0.339      0.229     1.48   0.139       -0.110       0.789
          2  |      0.960      0.229     4.19   0.000        0.511       1.410
             |
        drug |
   standard  |      0.000  (base)
        new  |     -0.055      0.241    -0.23   0.820       -0.527       0.418
             |
   time#drug |
      1#new  |      1.002      0.337     2.97   0.003        0.341       1.662
      2#new  |      2.097      0.390     5.38   0.000        1.333       2.862
             |
       _cons |      0.021      0.178     0.12   0.906       -0.327       0.369
------------------------------------------------------------------------------

Agresti reports the results with a single coefficient for time (0.48) and the time#drug (1.01) interaction which is interpreted as the new drug increasing the slope by 1.01 giving a faster rate of improvement.

My problem is how do I get a single coefficient rather than the individual main effects interactions or do I not need to worry about it. If so how do I interpret the results?

This is a subset of 100 results from the Agresti data which has 1020 results.

Thank you,
Julie

* Example generated by -dataex-. For more info, type help dataex
clear
input byte(diagnose drug) int id byte(time depression) double diag byte select
1 1   2 0 1 0 1
1 1   3 0 1 0 1
1 1   8 1 1 0 1
1 1   9 2 1 0 1
1 2  17 1 1 0 1
1 2  25 2 1 0 1
1 2  26 2 1 0 1
1 2  28 1 1 0 1
1 2  28 2 1 0 1
1 2  29 0 1 0 1
1 2  29 2 1 0 1
1 2  31 0 1 0 1
1 2  42 1 1 0 1
1 2  42 2 1 0 1
1 2  43 1 1 0 1
1 2  44 2 1 0 1
2 2  52 0 1 1 1
2 2  53 1 1 1 1
1 1  60 1 1 0 1
1 1  61 2 0 0 1
1 1  64 1 1 0 1
1 1  80 1 0 0 1
1 1  82 2 1 0 1
2 1  89 2 1 1 1
2 1  90 2 1 1 1
2 1  96 2 1 1 1
2 2  97 2 1 1 1
1 1 102 0 1 0 1
2 1 109 0 1 1 1
1 1 117 0 0 0 1
1 1 119 1 1 0 1
1 1 121 0 0 0 1
1 1 129 0 0 0 1
1 2 131 0 0 0 1
1 2 133 2 1 0 1
1 2 134 1 1 0 1
1 2 137 1 1 0 1
1 2 139 1 1 0 1
1 2 147 1 1 0 1
1 2 150 1 1 0 1
2 1 152 0 0 1 1
2 1 156 0 0 1 1
2 1 158 0 0 1 1
2 1 158 2 1 1 1
2 2 165 1 1 1 1
2 2 166 1 1 1 1
2 2 168 2 1 1 1
2 2 169 2 1 1 1
2 2 172 0 0 1 1
2 2 172 2 1 1 1
2 2 186 2 1 1 1
2 1 199 1 1 1 1
2 1 202 2 0 1 1
2 1 206 0 0 1 1
2 1 208 0 0 1 1
2 1 211 2 0 1 1
2 2 215 0 0 1 1
2 2 217 2 0 1 1
1 1 220 0 0 0 1
1 1 222 1 0 0 1
1 1 225 2 1 0 1
1 1 226 2 1 0 1
1 2 236 0 0 0 1
1 2 237 2 1 0 1
1 2 239 2 1 0 1
1 2 240 2 1 0 1
1 2 241 0 0 0 1
2 1 246 1 0 1 1
2 1 247 2 1 1 1
2 1 260 0 0 1 1
2 1 260 2 1 1 1
2 1 263 0 0 1 1
2 1 266 1 0 1 1
2 2 269 1 0 1 1
2 2 271 0 0 1 1
2 2 274 0 0 1 1
2 2 276 0 0 1 1
2 2 277 1 0 1 1
2 2 280 2 1 1 1
2 2 282 1 0 1 1
2 2 284 0 0 1 1
2 2 284 2 1 1 1
2 2 286 0 0 1 1
2 2 288 2 1 1 1
2 2 290 2 1 1 1
2 2 292 0 0 1 1
2 2 294 0 0 1 1
1 1 305 1 0 0 1
1 1 306 0 0 0 1
2 1 307 0 0 1 1
2 1 308 0 0 1 1
2 1 319 2 0 1 1
2 1 321 1 0 1 1
2 1 322 2 0 1 1
2 1 324 1 0 1 1
2 1 331 0 0 1 1
2 1 334 0 0 1 1
2 1 334 1 0 1 1
2 2 335 1 0 1 1
2 2 337 0 0 1 1
end
label values diagnose diagnose
label def diagnose 1 "mild", modify
label def diagnose 2 "severe", modify
label values drug drug
label def drug 1 "standard", modify
label def drug 2 "new", modify

------------------ copy up to and including the previous line ------------------

Listed 100 out of 1020 observations