为箱线图旋转数据集并计算变量的演变

Question

I am trying to code a plot with 3 boxplots, one per year (N2, N1 and N). I want to plot variable "DMS".

library(dplyr)
library(ggplot2)

mean_DMS_2021 <- NCN_dataset %>%
   filter(type_de_sejour == "Hospitalisé")%>%
   group_by(code_2)%>%
   summarise(mean_DMS_hosp_N = mean(as.numeric(DMS_hosp_N)))

mean_DMS_2020 <- NCN_dataset %>%
  filter(type_de_sejour == "Hospitalisé")%>%
  group_by(code_2)%>%
  summarise(mean_DMS_hosp_N1 = mean(as.numeric(DMS_hosp_N1)))

mean_DMS_2019 <- NCN_dataset %>%
  filter(type_de_sejour == "Hospitalisé")%>%
  group_by(code_2)%>%
  summarise(mean_DMS_hosp_N2 = mean(as.numeric(DMS_hosp_N2)))

mean_DMS_hosp_19_20 <- as.data.frame(merge(mean_DMS_2019, mean_DMS_2020))
mean_DMS_hosp <- as.data.frame(merge(mean_DMS_hosp_19_20, mean_DMS_2021))
View(mean_DMS_hosp)

# I here show the mean of DMS for each year, all specialties together. 

boxplot(mean_DMS_hosp$mean_DMS_hosp_N2)
boxplot(mean_DMS_hosp$mean_DMS_hosp_N1)
boxplot(mean_DMS_hosp$mean_DMS_hosp_N)

I tried to pivot my data because ideally I would like to plot, per year, a boxplot for variable DMS, in order to see the evolution throughout the years of the distribution of this variable, and then facet (or anything else to be able to show the differences between specialties) per specialty or per "nom": either keeping only the specialties with the highest variance between the observations for the variable DMS, or filtering to keep only the DMS that have evolved the most throughout the years (I guess I would have to create several variables of "evolution" showing the evolution of DMS between N2 and N1, between N1 and N, and the average annual growth rate between N2 and N)).

 library(data.table)

 NCN_dataset_long  <- melt(data.table::setDT(NCN_dataset), 
                          measure.vars=list(c(4,5,6), c(7,8,9) , c(10,11,12)),
                          variable.name='time_year', 
                      value.name=c('sejour', 'CA',"DMS"))[,
                                                          time_year:= 
 paste0('N',time_year)][order(nom,type_de_sejour, code_2, site)]

Here is the structure of my dataset, for reproducibility:

setDT(structure(list(nom = c("CHRISTOPHE", "CHRISTOPHE", 
"PABLO", "JEAN-MARC", "YVES", 
"GUILLAUME"), type_de_sejour = c("Ambulatoires", 
"Externes", "Ambulatoires", "Ambulatoires", "Ambulatoires", "Ambulatoires"
), code_2 = c("Ortho", "Ortho", "Neuro Chir", "Cardio", "Radio", 
"ARE"), sejours_N2 = c(1046, 0, 4, 6, 4, 4), sejours_N1 = c(1001, 
1, 77, 26, 9, 1), sejours_N = c(1078, 0, 115, 140, 9, 1), CA_N2 = c(609862, 
0, 2002, 3296, 1457, 1253), CA_N1 = c(597436, 24, 119573, 22098, 
3026, 322), CA_N = c(668426, 0, 196852, 134095, 3454, 345), DMS_hosp_N2 = c("0", 
"0", "0", "0", "0", "0"), DMS_hosp_N1 = c("0", "0", "0", "0", 
"0", "0"), DMS_hosp_N = c("0", "0", "0", "0", "0", "0"), site = c("PGS", 
"PGS", "FRA", "FRA", "PGS", "FRA")), row.names = c(NA, -6L), class = c("data.table", 
"data.frame")))

Answer 1

I'll use CA here since your DMS is all zero.

library(ggplot2)
library(tidyr) # pivot_longer
pivot_longer(NCN_dataset, -c(nom, type_de_sejour, code_2, site), names_pattern = "(.*)_(N.*)", names_to = c(".value", "time_year")) |>
  ggplot(aes(x=time_year, y=CA)) +
  geom_boxplot()

You control orientation by using either x= or y= aesthetics for the numeric property.