quartile plots

The goal of this excercise is to show how to use the cut function to create concentration bins. One common use is to create a quartile plot of the proportion of responders given a concentration range, thus the objective is to:

• create concentration bins
• calculate mean, upper confidence, and lower confidence intervals for proportion of responders in each bin
• plot these

library(PKPDdatasets)
dat <- pain_relief

library(ggplot2)
library(dplyr)

## 
## Attaching package: 'dplyr'
## 
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## 
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

dat %>% filter(TIME == 2.0) %>% ggplot(aes(x = CONC, y = PAINRELIEF)) + geom_point()

dat_2hr <- dat %>% filter(TIME ==2)

dat_2hr$CONC_CUT <- cut(dat_2hr$CONC, breaks = quantile(dat_2hr$CONC), include.lowest = TRUE)

# mean conc by cut
mean_conc <- dat_2hr %>% group_by(CONC_CUT) %>%
  mutate(BIN_CONC = mean(CONC),
         PROP_RELIEF = mean(PAINRELIEF), 
         n = n(),
         SE = sqrt((PROP_RELIEF*(1-PROP_RELIEF)/n)),
         UCI = PROP_RELIEF + 1.96*SE, 
         LCI = PROP_RELIEF - 1.96*SE) %>% 
                                                         select(CONC, PAINRELIEF, PAINSCORE, DOSE, CONC_CUT, BIN_CONC ,PROP_RELIEF, UCI, LCI)

# need mean 'bin' conc values to map to x axis for plotting numerical concentrations on the CONC

ggplot(mean_conc, aes(x = CONC, y = PAINRELIEF)) + geom_point() + geom_errorbar(aes(x = BIN_CONC, ymin = LCI, ymax = UCI)) + geom_point(aes(x = BIN_CONC, y = PROP_RELIEF), size = 2)