add_poisson_gest: Add binomial G-Estimate

In alexhallam/gest:

Description

adds an estimated shinkage value obtained by G-Modeling

Usage

add_poisson_gest(
  tbl,
  value,
  cred_level = 0.8,
  theta_from = -4,
  theta_to = 5,
  theta_by = 0.1,
  probs = c(0.005, 0.025, 0.165, 0.25, 0.5, 0.75, 0.835, 0.975, 0.995)
)

Arguments

tbl	a table which *requires* two columns. A column with the number of successes (x) and a column with totals (n).
cred_level	level of credible interval to compute. For example and interval of .80 will return a 20th percentile lower interval and a 80th percentile upper interval
theta_from	start of parameter space
theta_to	end of parameter space
theta_by	increment of parameter space
x	column with number of successes
n	column with totals

Value

The original table with the following columns

.gest_dist	Full posterior distribution. Grain may be increased with the theta_* parameters
.raw	Estimate (success / total)
.gest	Posterior EB g-estimate
.lo	Lower bound of credible interval for EB estimate
.hi	Upper bound of credible interval for EB estimate

Examples

dontrun{
#Example 1 ##################################
library(tidyverse)
library(ZIM)
data(syph)
df <- syph %>%
  pivot_longer(cols = starts_with("a"), names_to = "key", values_to = "value")

df %>%
  add_poisson_gest(value)
  }
#Example 2 ##################################
data("df")
df %>% add_poisson_gest() %>% tidyr::unnest(.probs)

alexhallam/gest documentation built on March 24, 2020, 1:22 a.m.