In oizin/funnelplot: Create Funnel Plots

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Using XGBoost in funnelplot

library(funnelplot)
library(xgboost)

data(example_data,package="funnelplot")
head(example_data)

The default is logistic regression:

flr <- funnelModel(test ~ gender + age | hosp_id, pointTarget(limits = c(0.05,0.01), standardised = TRUE), 
                  adj = adjParams(model = "logistic",method = "out_of_fold",nfolds = 5), 
                  data = example_data)
plot(flr)

By defining some functions we can use xgboost:

xg_fit <- function(formula,data) {
  xx <- model.matrix(formula,data=data)[,-1]  # remove intercept
  y <- data[[all.vars(formula)[1]]]
  out <- xgboost(data = xx, label = y, max.depth = 2, eta = 1, nthread = 2, nrounds = 2, objective = "binary:logistic")
  class(out) <- c("xg_fit",class(out))
  out
}

# this next step is only required due to the formula-only interface currently in use
predict.xg_fit <- function(object,newdata,formula,...) {
  xx <- model.matrix(formula,data=newdata)[,-1]  # remove intercept
  class(object) <- class(object)[-1]
  predict(object,xx)
}

fxg <- funnelModel(test ~ gender + age | hosp_id, pointTarget(limits = c(0.05,0.01), standardised = TRUE), 
                  adj = adjParams(model = xg_fit,method = "out_of_fold",nfolds = 5), 
                  data = example_data)
plot(fxg)