用應用族替換生成圖形的嵌套 for 循環

Question

背景

我有一個來自調查的定量數據集。 我想 plot 擬合我擁有的值的三角形分布（最小lb ，最大ub和模式ml ）。 請注意，我使用的是rtriang() ，因為我的數據不包含可以擬合密度 function 的分位數。 至少這是我的理解。

問題

• 我現在正在使用一個丑陋的嵌套 for 循環，使用 apply-family 函數執行此操作可能更有效，盡管我不知道如何執行此操作。 我該怎么做？

代碼

library(data.table)
library(ggplot2)
library(mc2d)

scenarios <- c("s1", "s2")
questions <- c("q1", "q2")
respondents <- c("1","2","3")

data_long <- data.frame(id=c("1","2","3", "1","2","3", "1","2","3",
                               "1","2","3", "1","2","3", "1","2","3",
                               "1","2","3", "1","2","3", "1","2","3",
                               "1","2","3", "1","2","3", "1","2","3"),
                         variable=c("s1_q1_ml", "s1_q1_ml", "s1_q1_ml",
                                      "s1_q1_lb", "s1_q1_lb", "s1_q1_lb",
                                      "s1_q1_ub", "s1_q1_ub", "s1_q1_ub",
                                      "s1_q2_ml", "s1_q2_ml", "s1_q2_ml",
                                      "s1_q2_lb", "s1_q2_lb", "s1_q2_lb",
                                      "s1_q2_ub", "s1_q2_ub", "s1_q2_ub",
                                      "s2_q1_ml", "s2_q1_ml", "s2_q1_ml",
                                      "s2_q1_lb", "s2_q1_lb", "s2_q1_lb",
                                      "s2_q1_ub", "s2_q1_ub", "s2_q1_ub",
                                      "s2_q2_ml", "s2_q2_ml", "s2_q1_ml",
                                      "s2_q2_lb", "s2_q2_lb", "s2_q1_lb",
                                      "s2_q2_ub", "s2_q2_ub", "s2_q1_ub"),
                         value=c(70, 70, 70, 60, 60, 60, 80, 80, 80,
                                   70, 70, 70, 60, 60, 60, 80, 80, 80,
                                   70, 70, 70, 60, 60, 60, 80, 80, 80,
                                   70, 70, 70, 60, 60, 60, 80, 80, 80))

data_long <- setDT(data_long)

for (i in respondents) {
  for (j in scenarios) {
    for (k in questions) {
      t <- rtriang(n =100000, min=as.numeric(data_long[id==i & variable == paste(j, k, "lb", sep = "_")]$value), 
                   mode=as.numeric(data_long[id==i & variable == paste(j,k, "ml", sep = "_")]$value),
                   max=as.numeric(data_long[id==i & variable == paste(j,k, "ub", sep = "_")]$value))

      # Displaying the samples in a density plot
      plot <- ggplot() + geom_density(aes(t)) + xlim(0,100) + xlab("Probability in %")
      ggsave(plot,filename=paste(i,j,k,".png",sep="_"))
    }
  }
}

Answer 1

一個tidyverse方法：

library(tidyverse)
library(mc2d)

all_plots <- data_long %>%
               separate(variable, c("scenarios", "questions", "temp"),
                         sep = "_") %>% 
               group_split(id, scenarios, questions) %>%
               map(~{
                    temp <- rtriang(
                      n =100000, 
                      min = .x %>% filter(temp == 'lb') %>% pull(value),
                      mode = .x %>% filter(temp == 'ml') %>% pull(value),
                      max = .x %>% filter(temp == 'ub') %>% pull(value))
                      plot <- ggplot() + 
                               geom_density(aes(temp)) + xlim(0,100) + 
                               xlab("Probability in %")
                       ggsave(filename = paste(.x$id[1],.x$scenarios[1],
                                        .x$questions[1],".png",sep="_"), plot)
                   })