I have been trying to calculate confidence intervals for binomial distributions through the Hmisc R package. Specifically, I used the binconf function which does its job perfectly.
library(plyr)
library(Hmisc)
Student <- c("A", "B", "C")
TP <- c(13, 36, 43)
obs.pos <- c(16, 37, 48)
df <- data.frame(Student, TP, obs.pos)
df1 <- df %>%
plyr::mutate(Sen = binconf(TP, obs.pos, alpha = 0.05, method = "wilson", return.df = TRUE))
df1 %>% View()
# Student TP obs.pos Sen.PointEst Sen.Lower Sen.Upper
#1 A 13 16 0.8125000 0.5699112 0.9340840
#2 B 36 37 0.9729730 0.8617593 0.9986137
#3 C 43 48 0.8958333 0.7783258 0.9546783
Unfortunately, I feel that the function creates a data frame within my original data frame and that does not allow me to apply basic functions on my output anymore. For instance, I cannot select columns (by using dplyr) or round digits because R is not able to find the created columns (such as Sen.PointEst, Sen.Lower, Sen.Upper). Below, the structure of my output.
df1 %>% str()
#'data.frame': 3 obs. of 4 variables:
# $ Student: Factor w/ 3 levels "A","B","C": 1 2 3
# $ TP : num 13 36 43
# $ obs.pos: num 16 37 48
# $ Sen :'data.frame': 3 obs. of 3 variables:
# ..$ PointEst: num 0.812 0.973 0.896
# ..$ Lower : num 0.57 0.862 0.778
# ..$ Upper : num 0.934 0.999 0.955
I would like to have all the columns at the first level of my output so that I can easily apply all the normal functions to my output.
Thanks for any help!
We have a column that is data.frame inside a data.frame. One option to flatten out the data.frame will be to call data.frame within do.call
dfN <- do.call(data.frame, df1)
Or another option is to call the binconf within do
df %>%
do(data.frame(., Sen = binconf(.$TP, .$obs.pos, alpha = 0.05, method = "wilson")))