benchmark vs lme4

[bdilday]

For a binomial outcome we can compare execution time to lme4::glmer. These are the current results on an example problem.

microbenchmark::microbenchmark(mnre_mod = mnre_fit(y ~ 1 + (1|fct01) + (1|fct02), data=ev$fr, verbose=0), glmer_mod <- glmer(ev$frm, data=ev$fr, family='binomial', nAGQ=0), times = 5)
Unit: milliseconds
                                                                         expr       min        lq
                                                                     mnre_mod 9442.0576 9480.7341
 glmer_mod <- glmer(ev$frm, data = ev$fr, family = "binomial",      nAGQ = 0)  893.5765  904.6866
      mean    median        uq      max neval
 9532.4786 9505.4875 9606.3683 9627.746     5
  981.5733  921.1398  959.1317 1229.332     5

[bdilday]

The pull request #9 gives a slight speed up.

microbenchmark::microbenchmark(mnre_mod  = mnre_fit(ev$frm, data=ev$fr),  glmer_mod <- glmer(ev$frm, data=ev$fr, family='binomial', nAGQ=0), times=5)
Unit: milliseconds
                                                                         expr       min        lq      mean   median
                                                                     mnre_mod 5407.3584 5499.9719 6814.7301 6045.798
 glmer_mod <- glmer(ev$frm, data = ev$fr, family = "binomial",      nAGQ = 0)  887.3456  890.4083  992.7495  905.885
        uq      max neval
 8227.8819 8892.640     5
  906.0734 1374.035     5