Question

我在2 - 4年内观察过a - d个主题，每年都会报告一个数值。我想提取每个主题的第一个和最后一个值，忽略NA。如何创建新变量first_value和last_value？在这个例子中，我已经包含了所需的结果：

df <- data.frame(subject = c("a","b","c","d"), 
                 year1 = c(1, 2, NA, NA),
                 year2 = c(3, 4, NA, 5),
                 year3 = c(6, 7, 8, NA),
                 year4 = c(9, 10, NA, 11),
                 first_value <- c(1, 2, 8, 5),
                 last_value <- c(9, 10, 8, 11))

如果变量year1 - year4是绝对的，会有什么解决方案？

Answer 1

使用data.table包：

library(data.table)
setDT(df)[, `:=` (first_value = na.omit(unlist(.SD))[1], 
                  last_value = tail(na.omit(unlist(.SD)),1)), 
          by = subject][]

给出：

   subject year1 year2 year3 year4 first_value last_value
1:       a     1     3     6     9           1          9
2:       b     2     4     7    10           2         10
3:       c    NA    NA     8    NA           8          8
4:       d    NA     5    NA    11           5         11

根据@alexis_laz的建议，您可以按如下方式使用max.col来获取相关值：

f <- max.col(!is.na(df[c("year1", "year2", "year3", "year4")]), 'first')
l <- max.col(!is.na(df[c("year1", "year2", "year3", "year4")]), 'last')

df$first_value <- sapply(seq_along(f), function(i) df[,-1][i,f[i]])
df$last_value <- sapply(seq_along(l), function(i) df[,-1][i,l[i]])

这会得到相同的结果。正如@alexis_laz在评论中所建议的那样，可以进一步改进：

m <- as.matrix(df[c("year1", "year2", "year3", "year4")])

f <- max.col(!is.na(m), 'first')
l <- max.col(!is.na(m), 'last')

df$first_value <- df[-1][cbind(1:nrow(df), f)]
df$last_value <- df[-1][cbind(1:nrow(df), l)]

并使用dplyr和tidyr套餐：

library(dplyr)
library(tidyr)

df %>% 
  gather(year, val, 2:5) %>% 
  filter(!is.na(val)) %>% 
  group_by(subject) %>% 
  summarise(first_value = first(val),
            last_value = last(val)) %>% 
  left_join(df, ., by = 'subject')

警告：不使用filter并在na.omit(val)中使用val[!is.na(val)]（或summarise）的变体：

df %>% 
  gather(year, val, 2:5) %>% 
  group_by(subject) %>% 
  summarise(first_value = first(na.omit(val)),
            last_value = last(na.omit(val))) %>% 
  left_join(df, ., by = 'subject')

由于报告的错误here和here，

无效。

Answer 2

使用data.frame和gather

#Used packages
library(tidyr)
library(dplyr)

subject<-c("a","b","c","d")  
year1 <- c(1, 2, NA, NA)  
year2 <- c(3, 4, NA, 5)  
year3 <- c(6, 7, 8, NA)  
year4 <- c(9, 10, NA, 11)  

dt = data.frame(subject, year1, year2, year3, year4)

gather()将多列折叠为一列

dt <- dt %>% gather(year, value, year1:year4)

summarise( )：执行所选变量的摘要统计

dt %>% group_by(subject)%>% 
        summarise(first_value = min(value, na.rm=TRUE),
                  last_value = max(value, na.rm=TRUE))

输出：

# A tibble: 4 × 3
  subject first_value last_value
   <fctr>       <dbl>      <dbl>
1       a           1          9
2       b           2         10
3       c           8          8
4       d           5         11

从数据框中的多个列中提取第一个和最后一个值

2 个答案: