按日期选择最近一年和最完整的行 R

Question

我有一个时间序列数据，我试图在其中获取每个 ID 的最新有效响应。这是我的起始测试数据集：

dta <- data.frame(
  uniqueID = c("950513","950513", "950513","951634","951634", "951634","951640","951640", "951640",
               "951641","951641", "951641","951646","951646", "951646","952732","952732", "952732",
               "952895", "952895", "952895","952909","952909", "952909","952910","952910", "952910",
               "952911","952911", "952911"),
  year = c(2014, 2015, 2016, 2014, 2015, 2016, 2014, 2015, 2016, 2014, 2015, 2016, 2014, 2015, 2016,
           2014, 2015, 2016, 2014, 2015, 2016, 2014, 2015, 2016, 2014, 2015, 2016, 2014, 2015, 2016),
  var1 = c(2115, NA, 2115,NA, NA, NA, NA, 2903, NA, NA, 6040, 6040, 6103, 6103, 6268, 6511, 7101, 7101,
           7105, 7105, 7105,8901, 8520, 8901, 8540, 8540, NA, 8854, 2108, NA),
  var2 = c(2115, "", 2115,"", "", "", "", 2903, "", "", 6040, 6040, 6103, 6103, 6268, 6511, 7101, 7101,
           7105, 7105, 7105,8901, 8520, 8901, 8540, 8540, "", 8854, 2108, ""),
  var3 = c("turtle", "turtle", "turtle","", "tortoise","", "turtle", "", "", "", "turtle", "tortoise", 
           "tortoise", "tortoise", "", "", "turtle", "", "turtle", "tortoise", "turtle", "tortoise", "", "",
           "", "", "", "tortoise", "turtle", "tortoise"),
  var4 = c("", "", "", "", "", "", "", "yellow", "", "blue", "", "", "", "", "", "", "", "red", "red", "", "",
           "", "", "", "pink", "", "", "", "", "")
)

我希望每个 uniqueID 都有一个，我将在其中获得最新（按年份）的响应，其中包含所有列 var1-var4 的该 ID 的最完整信息。一步一步，这是我想做的：

• 每个唯一ID 有一行输出
• 默认情况下，选择最近一年（行）的响应（响应在 var1 - var4 中）
• 如果最近一年的响应 (var1-var4) 有缺失值或空白，请忽略该 ID 的同一列中的不同值，必要时折叠年份。例如2014&2015

根据我的测试数据集，我希望数据集最终看起来像这样：

result <- data.frame(
  uniqueID = c("950513","951634","951640","951641","951646","952732","952895","952909","952910",
                       "952911"),
year = c("2016","2015","2015&2014","2016&2014", "2016&2015", "2016&2015", "2016&2014", "2016&2014", "2015&2014", "2016&2015" ),
var1 = c(2115,NA,2903, 6040, 6268, 7101, 7105, 8901, 8540, 2108),
var2 = c(2115,"", 2903, 6040, 6268, 7101, 7105, 8901, 8540, 2018),
var3 = c("turtle","tortoise", "turtle", "tortoise", "tortoise", "turtle", "turtle", "tortoise","", "tortoise"),
var4 = c("","", "yellow", "blue", "", "red", "red", "", "pink", "")
)

或者，年份可以拆分为多个列 year1、year2。

我使用 data.table 尝试了几种不同的解决方案，但这些解决方案通常专注于最近一年或最完整的案例，但不是两者兼而有之。如果需要，很乐意提供更多说明。

Answer 1

应该这样做。

is_valid <- function(x){
  !(x == "" | is.na(x))
}
valid_x <- function(x){
  if(any(is_valid(x))){
    x[which(is_valid(x))]
  }
  else{
    ""
  }
}
result <- dta %>% 
  mutate(across(contains("var"), as.character)) %>%
  pivot_longer(contains("var"), names_to="var", values_to="val") %>% 
  group_by(uniqueID, var) %>% 
  mutate(last_val = last(valid_x(val)), 
         last_year = max(year[which(val == last_val)])) %>%
  filter(year == last_year) %>% 
  ungroup %>% 
  group_by(uniqueID) %>% 
  mutate(last_year = paste(sort(unique(last_year[which(is_valid(last_val))]), decreasing=TRUE), collapse="&")) %>%
  select(-c(val, year)) %>% 
  pivot_wider(names_from="var", values_from="last_val") %>% 
  rename("year" = "last_year")

result
# # A tibble: 10 x 6
# # Groups:   uniqueID [10]
#   uniqueID year      var1  var2   var3       var4    
#   <chr>    <chr>     <chr> <chr>  <chr>      <chr>   
# 1 950513   2016      2115  "2115" "turtle"   ""      
# 2 951634   2015      NA    ""     "tortoise" ""      
# 3 951640   2015&2014 2903  "2903" "turtle"   "yellow"
# 4 951641   2016&2014 6040  "6040" "tortoise" "blue"  
# 5 951646   2016&2015 6268  "6268" "tortoise" ""      
# 6 952732   2016&2015 7101  "7101" "turtle"   "red"   
# 7 952895   2016&2014 7105  "7105" "turtle"   "red"   
# 8 952909   2016&2014 8901  "8901" "tortoise" ""      
# 9 952910   2015&2014 8540  "8540" ""         "pink"  
# 10 952911   2016&2015 2108  "2108" "tortoise" ""