Sabre

EPILEP [epilep.dat]
------
Univariate Poisson model
236 observations, 59 cases, 6+1 (fixed+random) parameters

SABRE: 64-point quadrature, accurate arithmetic [log likelihood = -665.519]
STATA: xtpois, 64-point non-adaptive quadrature [log likelihood = -665.438]
GLLAMM: 64-point non-adaptive quadrature [log likelihood as STATA]
IGLS: approximate timing due to running in interactive mode
MCMC: default settings (Gamma priors), 25000 iterations;
      approximate timing due to running in interactive mode

Note that SABRE finds a 'better' solution [log likelihood = -665.233] from a
starting value of sigma = 0.52, and this same solution can be found in STATA
from a starting value of /lnsig2u = -1.3 and in GLLAMM starting from the
associated parameter estimates to 2dp.

         SABRE           STATA          GLLAMM            IGLS            MCMC
lbas     0.922 (0.096)   0.919 (0.097)  as STATA         0.880 (0.137)   0.871 (0.128)
treat   -0.723 (0.325)  -1.028 (0.333)                  -0.887 (0.415)  -0.995 (0.451)
lbas_tr  0.200 (0.131)   0.391 (0.131)                   0.342 (0.211)   0.364 (0.222)
lage     0.476 (0.379)   0.500 (0.417)                   0.507 (0.357)   0.454 (0.367)
visit   -0.294 (0.101)  -0.294 (0.101)                  -0.294 (0.104)  -0.293 (0.101)
const    2.002 (0.183)   2.177 (0.184)                   2.218 (0.228)   2.143 (0.251)
sigma    0.508 (0.052)   0.524 (0.052)                   0.530 (0.058)   0.546 (0.068)

time    00'06"          00'01"          00'18"          00'00"          00'25"