API

blb_one_subset(rng, m; N, subset_size, n_boots, method, solver, verbose, nonparametric_boot)

Performs Bag of Little Bootstraps on a subset.

Positional arguments

• rng: random number generator. Default to the global rng.
• m: an object of type blblmmModel or WSVarLmmModel

Keyword arguments

• N: number of individuals/clusters in the full dataset.
• subset_size: number of individuals/clusters in each subset.
• n_boots: number of bootstrap iterations. Default to 1000.
• method: a symbol, either :ML or :WiSER.
• solver: solver for the optimization problem.
• verbose: Bool, whether to print bootstrap progress.
• nonparametric_boot: Bool, whether to use nonparametric bootstrap. For WiSER models, only nonparametric bootstrap is supported.

Values

• subset_estimates: an object of type SubsetEstimates

blb_full_data(rng, datatable; feformula, reformula, wsvarformula, id_name, cat_names, subset_size, n_subsets, n_boots, method, solver, verbose,  nonparametric_boot)

Performs Bag of Little Bootstraps on the full dataset

Positional arguments

• rng: random number generator. Default to the global rng.
• datatable: a data table type that is compatible with Tables.jl.

Keyword arguments

• feformula: model formula for the fixed effects.
• reformula: model formula for the random effects.
• wsvarformula: model formula for the fixed effects of the within-subject variance. For linear mixed models, it should be @formula(y ~ 1). Only need to be specified when method = :WiSER.
• id_name: name of the cluster identifier variable. String.
• cat_names: a vector of the names of the categorical variables.
• subset_size: number of clusters in the subset.
• n_subsets: number of subsets.
• n_boots: number of bootstrap iterations. Default to 1000
• solver: solver for the optimization problem.
• method: fitting the model by maximum-likelihood (:ML) or GEE (:WiSER)
• verbose: Bool, whether to print bootstrap progress (percentage completion)
• nonparametric_boot: Bool, whether to use Nonparametric bootstrap

Values

• result: an object of the blbEstimates type

blb_db(rng, con, table_name; feformula, reformula, id_name, cat_names, subset_size, n_subsets, n_boots, solver, verbose,  nonparametric_boot)

Performs Bag of Little Bootstraps on databases.

Positional arguments

• rng: random number generator. Default to the global rng.
• con: an object of type MySQL.Connection created by the function DBInterface.connect.
• table_name: table name for the longitudinal data.

Keyword arguments

• feformula: model formula for the fixed effects.
• reformula: model formula for the random effects.
• id_name: name of the cluster identifier variable. String.
• cat_names: a vector of the names of the categorical variables.
• subset_size: number of clusters in the subset.
• n_subsets: number of subsets.
• n_boots: number of bootstrap iterations. Default to 1000
• solver: solver for the optimization problem.
• verbose: Bool, whether to print bootstrap progress (percentage completion)
• nonparametric_boot: Bool, whether to use Nonparametric bootstrap

Values

• result: an object of the blbEstimates type

SubsetEstimates

BLB linear mixed model estimates from one subset.

blbEstimates

BLB linear mixed model estimates, which contains blb parameters and a vector of SubsetEstimates.

confint(subset_ests, level)

Calculate confidence intervals using estimates from one subset.

Positional arguments

• subset_ests: an object of type SubsetEstimates
• level: confidence level, usually set to 0.95

confint(blb_ests, level)

Calculate confidence intervals using estimates from all subsets.

Positional arguments

• blb_ests: an object of type blbEstimates
• level: confidence level, usually set to 0.95