将因子类别从一列合并为新列

Question

这是我第一次发问。我对R很陌生，我寻找答案已经有一段时间了，却没有找到答案。就这么办了。我有一个非常大的数据集(超过140 K的数据集)，其中一列包含"programtype“的类别，选项如下：

• 联邦机构
• 联邦机构/大学
• 国家调查方案
• LTER
• 大学
• 非营利机构
• 国家局
• 国家机构/公民监测方案
• 国家机构/大学/公民监测方案
• 部落机构

我想要做的是创建一个新的列，在其中我合并一些类别在一起。我想：

• 联邦机构、联邦机构/大学、国家调查计划改为联邦机构/大学
• 大学改为 LTER /
• 非营利机构转换为非盈利机构
• 国家局改为国事局
• 国家机构/公民监测方案、国家机构/大学/公民监测方案转化为公民科学监测方案
• 部落机构改为部落机构

他们中的一些会保持原样。我尝试过ifelse语句，但似乎很难确定原始专栏中的内容，并返回NAs进行大量的观察。我检查了我所有的拼写，所以不是这样的。这是我在这里的另一个答案的基础上尝试的。我的数据集的名称是TP_state，其他列的名称是lagoslakeid。但是，它不能正常工作。任何帮助都将不胜感激！

x <- c(TP_state$programtype) 
y <- c(TP_state$lagoslakeid)
df <- data.frame(x,y)
DT <- data.table(df)
DT[, Program_Type := ifelse(x %in% c("Federal Agency", "Federal Agency/University", "National Survey Program"), "Federal Agency/University",
                 ifelse(x %in% c("LTER", "University"), "LTER/University",
                 ifelse(x %in% c("Non-Profit Agency"), "Non-Profit Agency",       
                 ifelse(x %in% c("State Agency"), "State Agency",
                 ifelse(x %in% c("State Agency/University/Citizen Monitoring Program", "State Agency/Citizen Monitoring Program"), "Citizen Monitoring Program", 
                 ifelse(x %in% c("Tribal Agency"), "Tribal Agency", NA))))))]  

Answer 1

我会尝试这样的方法。请告诉我它是否对你有用！

for(i in 1:length(df$column_with_factors)){
  if(grepl(pattern = 'federal agency|national survey program', x = df$column[i], ignore.case = TRUE)){
    x <- 'Federal Agency/University'
  } else if(grepl(pattern = '^lter$|^university$', x = df$column[i], ignore.case = TRUE)){
    x <- 'LTER/University'
  } else if(grepl(pattern = 'non-profit agency', x = df$column[i], ignore.case = TRUE)){
    x <- 'Non-profit Agency'
  } else if(grepl(pattern = '^state agency$', x = df$column[i], ignore.case = TRUE)){
    x <- 'State Agency'
  } else if(grepl(pattern = 'state agency/(citizen monitoring program|university/citizen monitoring program)', x = df$column[i], ignore.case = TRUE)){
    x <- 'Citizen Science Monitoring Program'
  } else if(grepl(pattern = 'tribal agency', x = df$column[i], ignore.case = TRUE)){
    x <- 'Tribal Agency'
  } else x <- NA
}

df$column_with_factors <- as.factor(df$column_with_factors)

但这会跑得更快：

df$column_with_factors <- sapply(df$column_with_factors, function(x){
  if(grepl(pattern = 'federal agency|national survey program', x = x, ignore.case = TRUE)){
    x <- 'Federal Agency/University'
  } else if(grepl(pattern = '^lter$|^university$', x = x, ignore.case = TRUE)){
    x <- 'LTER/University'
  } else if(grepl(pattern = 'non-profit agency', x = x, ignore.case = TRUE)){
    x <- 'Non-profit Agency'
  } else if(grepl(pattern = '^state agency$', x = x, ignore.case = TRUE)){
    x <- 'State Agency'
  } else if(grepl(pattern = 'state agency/(citizen monitoring program|university/citizen monitoring program)', x = x, ignore.case = TRUE)){
    x <- 'Citizen Science Monitoring Program'
  } else if(grepl(pattern = 'tribal agency', x = x, ignore.case = TRUE)){
    x <- 'Tribal Agency'
  } else x <- NA
})

df$column_with_factors <- as.factor(df$column_with_factors)

Answer 2

forcats包在对这样的任务进行重新编码方面非常出色。

首先创建一些演示数据..。

library(tidyverse)
library(forcats)

df <-
  tibble(
    programtype = c(
      "Federal Agency",
      "Federal Agency",
      "Federal Agency",
      "State Agency/University/Citizen Monitoring",
      "State Agency/University/Citizen Monitoring Program",
      "Federal Agency/University",
      "National Survey Program",
      "LTER",
      "University",
      "Non-Profit Agency",
      "Non-Profit Agency",
      "Non-Profit Agency",
      "Non-Profit Agency",
      "Non-Profit Agency",
      "State Agency",
      "State Agency",
      "State Agency/Citizen Monitoring Program",
      "State Agency/University/Citizen Monitoring Program",
      "Tribal Agency",
      "Tribal Agency",
      "Tribal Agency"
    ),
    ID = 1:21
  )

然后使用fct_recode替换值..。

df %>%
  mutate(
    new_categories = fct_recode(
      programtype,
      "Federal Agency/University" = "Federal Agency",
      "Federal Agency/University" = "Federal Agency/University",
      "Federal Agency/University" = "National Survey Program",
      "LTER/University" = "LTER",
      "LTER/University" = "University",
      "Citizen Science Monitoring Program" = "State Agency/Citizen Monitoring Program",
      "Citizen Science Monitoring Program" = "State Agency/University/Citizen Monitoring"
    )
  )