R语言使用dcast统计时如何为因子变量的0观测水平显示0值

解决方法 #

1. 首先确认已为数据框的AEOUT列设置了完整的预设因子水平，你的初始数据没有应用这层设置，需要先运行：

df$AEOUT <- factor(df$AEOUT, levels = c("RECOVERED/RESOLVED","RECOVERED/RESOLVED WITH SEQUELAE", "RECOVERING/RESOLVING", "NOT RECOVERED/NOT RESOLVED", "FATAL", "UNKNOWN"))

2. 使用dcast时添加drop = FALSE参数，同时指定计数的聚合函数，空水平会自动填充为0，同时保留原有因子顺序：

library(reshape2)
library(dplyr)

test <- df %>% 
  dcast(AEOUT ~ AEREL1S, 
        fun.aggregate = length, 
        drop = FALSE,
        value.var = "AEREL1S")

如果你习惯使用tidyverse系列的函数，也可以用count+pivot_wider实现相同效果：

library(dplyr)
library(tidyr)

test <- df %>%
  count(AEOUT, AEREL1S, .drop = FALSE) %>%
  pivot_wider(
    names_from = AEREL1S,
    values_from = n,
    values_fill = 0
  )

内容的提问来源于stack exchange，提问作者Stataq