boot_tmeans: Bootstrap distribution of trial means

In femiguez/predintma: Prediction Intervals for Meta-Analysis

Description

Bootstrapped distribution of trial means

Usage

boot_tmeans(
  formula = NULL,
  data,
  R = 500,
  var.names = c("y", "trial"),
  ci.type = "basic",
  level = 0.95,
  ncpus = 1
)

Arguments

formula	standard R formula of the form response ~ trial
data	data frame
R	number of bootstrap samples
var.names	alternative to formula
ci.type	type of confidence interval (see boot::boot.ci)
level	prediction level
ncpus	number of cpus to use, passed to boot, negligible benefit for small datasets

Value

list with 'dat': data frame with the bootstrapped means,
'boot.ci': confidence interval computed by 'boot.ci'
and 'pdi': prediction interval based on quantiles

Examples

## Not run: 
require(ggplot2)
data(soyrs)
## Simply calculate the trial means
tmns <- aggregate(lrr ~ Trial_ID, data = soyrs, FUN = mean)
## Bootstrapped stratified trial means
btm <- boot_tmeans(lrr ~ Trial_ID, data = soyrs, R = 2e3)

pdi.cf <- pred_int_conformal_df(formula = lrr ~ Trial_ID, x = soyrs)

btmd <- btm$dat
btm.q <- btm$pdi

ggplot() + xlab("lrr") + 
geom_density(data = btmd, aes(x = ys)) + 
geom_jitter(data = soyrs, aes(x = lrr, y = 2.5)) + 
geom_jitter(data = tmns, aes(x = lrr, y = 5), color = "blue", size = 1.2) + 
geom_point(aes(x = pdi.cf[1], y = -1), color = "orange", size = 1.2) + 
geom_errorbarh(mapping = aes(xmin = pdi.cf[2], xmax = pdi.cf[3],
                              y = -1), color = "orange", size = 1.2) +
geom_point(aes(x = btm.q[1], y = -2), color = "red", size = 1.2) + 
geom_errorbarh(aes(xmin = btm.q[2], xmax = btm.q[3], y = -2),
               color = "red", size = 1.2)


## End(Not run)

femiguez/predintma documentation built on July 5, 2021, 4:16 a.m.