brrr: Improved and computationally stable relative risk estimation...

In Francesco16p/brrr: Accurate estimation in relative risk regression

Improved and computationally stable relative risk estimation for a binary exposure

Description

Improved and computationally stable relative risk estimation for a binary exposure

Usage

brrr(
  x,
  z = NULL,
  y,
  t,
  maxit = 150,
  start = NULL,
  param = "Richardson",
  method = "Mle"
)

Arguments

x	matrix of explanatory variables for the nuisance parameter, the first column is the intercept of the model
z	matrix of explanatory variables for log-relative risk, the first column is the intercept of the model
y	binary response variable
t	binary risk factor
maxit	maximum number of iteration for the estimation algorithm
start	initial guess for the parameter. If null vector of zeros
param	nuisance parameter. Possible values are "Richardson" or "Alternative"
method	estimation method. Possible values are "Mle", "MeanBr" and "MedianBr"

Examples

#####################################################
## Relative risk regression with brrr            ####
#####################################################

library(brrr)
library(HSAUR3)

##################################################
###       Respiratory dataset                  ###
### available on HSAUR3 R package              ###
##################################################

data <- HSAUR3::respiratory

## Creating response variable, binary risk factor and model matrices
id <- seq.int(from = 5, to =  555, by  = 5)
data_matrix <- model.matrix(~., data[id, c(1,2,3,4,5)])
y <- data_matrix[,6]
t <- data_matrix[,3]
x <- z <- data_matrix[,-c(3,6)]

## maximum likelihood fit using R package brrr, model: Richardson et al. (2017)
rr_mle <- brrr(x,z,y,t)
summary(rr_mle)

## Mean bias-reduced fit using the R package brrr, model: Richardson et al. (2017)
rr_meanBR <- brrr(x,z,y,t, method = "MeanBr")
summary(rr_meanBR)

## Median bias-reduced fit using the R function brrr, model: Richardson et al. (2017)
rr_medianBR <- brrr(x,z,y,t, method = "MedianBr")
summary(rr_medianBR)

## maximum likelihood fit using R package brrr, model: Alternative
rr_mleA <- brrr(x,z,y,t, param = "Alternative")
summary(rr_mleA)

## Mean bias-reduced fit using the R package brrr, model: Alternative
rr_meanBRA <- brrr(x,z,y,t, method = "MeanBr",param = "Alternative")
summary(rr_meanBRA)

## Median bias-reduced fit using the R function brrr, model: Alternative
rr_medianBRA <- brrr(x,z,y,t, method = "MedianBr",param = "Alternative")
summary(rr_medianBRA)

Francesco16p/brrr documentation built on Jan. 29, 2023, 2:07 a.m.