Parallelize 'fwb' functions

Henrik Bengtsson

+ =

The futurize package allows you to easily turn sequential code into parallel code by piping the sequential code to the futurize() function. Easy!

TL;DR

library(futurize)
plan(multisession)
library(fwb)

set.seed(123)
lm_fit <- lm(mpg ~ wt + am, data = mtcars)
b <- fwb(mtcars, statistic = function(data, w) {
  fit <- lm(mpg ~ wt + am, data = data, weights = w)
  coef(fit)
}, R = 999) |> futurize()

Introduction

This vignette demonstrates how to use this approach to parallelize fwb functions such as fwb() and vcovFWB().

The fwb package implements the fractional weighted bootstrap (also known as the Bayesian bootstrap). Rather than resampling units to include in bootstrap samples, random weights are drawn and applied to a weighted estimator. Given the resampling nature of bootstrapping, the algorithm is an excellent candidate for parallelization.

Example: Fractional weighted bootstrap

The fwb() function produces fractional weighted bootstrap samples of a statistic applied to data. For example, consider bootstrapping a linear model on the mtcars dataset:

library(fwb)

## Draw 999 bootstrap samples of the regression coefficients
set.seed(123)
b <- fwb(mtcars, statistic = function(data, w) {
  fit <- lm(mpg ~ wt + am, data = data, weights = w)
  coef(fit)
}, R = 999)

Here fwb() evaluates sequentially, but we can easily make it evaluate in parallel by piping to futurize():

library(fwb)
library(futurize)

set.seed(123)
b <- fwb(mtcars, statistic = function(data, w) {
  fit <- lm(mpg ~ wt + am, data = data, weights = w)
  coef(fit)
}, R = 999) |> futurize()

This will distribute the 999 bootstrap samples across the available parallel workers, given that we have set up parallel workers, e.g.

plan(multisession)

The built-in multisession backend parallelizes on your local computer and works on all operating systems. There are other parallel backends to choose from, including alternatives to parallelize locally as well as distributed across remote machines, e.g.

plan(future.mirai::mirai_multisession)

and

plan(future.batchtools::batchtools_slurm)

Example: Bootstrap variance-covariance matrix

The vcovFWB() function computes a bootstrap variance-covariance matrix for model coefficients:

library(futurize)
plan(multisession)
library(fwb)

lm_fit <- lm(mpg ~ wt + am, data = mtcars)
V <- vcovFWB(lm_fit, R = 999) |> futurize()

Supported Functions

The following fwb functions are supported by futurize():

• fwb() with seed = TRUE as the default
• vcovFWB() with seed = TRUE as the default