Chapter 16, Section 16.5

 

Crossover Trial on Cerebrovascular Deficiency

Marginal Logistic Regression Model

 

use "ecg.dta"

 

tsset id period

       panel variable:  id (strongly balanced)

        time variable:  period, 0 to 1

                delta:  1 unit

 

xtgee y trt period, family(binomial) link(logit) corr(unstr) robust 

 

Iteration 1: tolerance = .01591855

Iteration 2: tolerance = .00048302

Iteration 3: tolerance = .00002151

Iteration 4: tolerance = 6.380e-07

 

 

GEE population-averaged model                   Number of obs      =       134

Group and time vars:             id period      Number of groups   =        67

Link:                                logit      Obs per group: min =         2

Family:                           binomial                     avg =       2.0

Correlation:                  unstructured                     max =         2

                                                Wald chi2(2)       =      8.26

Scale parameter:                         1      Prob > chi2        =    0.0161

 

                                     (Std. Err. adjusted for clustering on id)

------------------------------------------------------------------------------

             |             Semirobust

           y |      Coef.   StdErr.      z    P>|z|     [95% Conf. Interval]

-------------+----------------------------------------------------------------

         trt |   .5688666   .2344814     2.43   0.015     .1092915    1.028442

      period |   .2950148   .2328697     1.27   0.205    -.1614015    .7514311

       _cons |    -1.2348   .3016183    -4.09   0.000    -1.825961   -.6436393

------------------------------------------------------------------------------

 

xtcorr

 

Estimated within-id correlation matrix R:

 

        c1      c2

r1  1.0000

r2  0.6243  1.0000

 

 

 

 

Mixed Effects Logistic Regression Model (Random Intercept)

 

.  xtmelogit y trt period || id:, intpoints(100)                            

 

Refining starting values: 

 

Iteration 0:   log likelihood = -76.707955  

Iteration 1:   log likelihood = -68.780976  

Iteration 2:   log likelihood = -68.316879  

 

Performing gradient-based optimization: 

 

Iteration 0:   log likelihood = -68.316879  

Iteration 1:   log likelihood = -68.130264  

Iteration 2:   log likelihood = -68.130169  

Iteration 3:   log likelihood = -68.130169  

 

Mixed-effects logistic regression               Number of obs      =       134

Group variable: id                              Number of groups   =        67

 

                                                Obs per group: min =         2

                                                               avg =       2.0

                                                               max =         2

 

Integration points = 100                        Wald chi2(2)       =      4.25

Log likelihood = -68.130169                     Prob > chi2        =    0.1193

 

------------------------------------------------------------------------------

           y |      Coef.   StdErr.      z    P>|z|     [95% Conf. Interval]

-------------+----------------------------------------------------------------

         trt |   1.863051   .9269169     2.01   0.044     .0463269    3.679775

      period |   1.037561   .8188658     1.27   0.205    -.5673866    2.642508

       _cons |  -4.081569   1.671067    -2.44   0.015      -7.3568   -.8063381

------------------------------------------------------------------------------

 

------------------------------------------------------------------------------

  Random-effects Parameters  |   Estimate   Std. Err.     [95% Conf. Interval]

-----------------------------+------------------------------------------------

id: Identity                 |

                   sd(_cons) |   4.943218   1.906526      2.321207    10.52702

------------------------------------------------------------------------------

LR test vs. logistic regression: chibar2(01) =    27.63 Prob>=chibar2 = 0.0000