Bootstrap details
All models are bootstrapped using
parameters::bootstap_model() function, which in turn
employs boot:boot() for the linear and generalized models,
and lme4::bootMer() for the mixed and generalized
mixed.
For linear and generalized models a nonparameteric bootstrap method is used. For each bootstrap estimate, data are sampled with replacement, a new model is estimated, and the parameters are saved. Confidence intervals are then constructed for the resulting distributions of estimates. The number of repetitions (bootstrap samples) can be set in | Options panel, Bootstrap rep., default is 1000. Confidence intervals can be constructed using the bootstrap percentile method (Bootstrap (percent)) or adjusted, bias corrected and accelerated bootstrap method, (Bootstrap (bca)).
For mixed and generalized mixed models a parametric bootstrap method is used. Each simulation generates new values of both the random effects and errors, using a normal distribution, with fixed parameters corresponding to the fitted model. The number of repetitions (bootstrap models) can be set in | Options panel, Bootstrap rep., default is 1000. Confidence intervals can be constructed using the bootstrap percentile method (Bootstrap (percent)).
In linux and MacOS, parallel computations is used with
multicore strategy. In Windows parallel computation is not
used.