将具有多个日期列的时间序列数据集转换为具有唯一日期列的时间序列或转换为动物园对象

Question

我有一个由多个不规则时间序列组成的大型数据集，每个序列都有一个特定的日期列。我想将此数据集转换为具有唯一日期列的数据框或转换为动物园对象。

我试过 read_xls(), read.zoo()。我试图用 pivot_longer() 重塑。我在网上搜索，但我还没有找到任何解决方案。

date1 Var1 date2 Var2 date3 Var3
2023-01-13 100.1 2023-01-11 99.7 2022-11-24 102.3
2023-01-16 104.5 2023-01-12 NA 2022-11-25 99.9
2023-01-17 101.6 2023-01-13 99.9 2022-11-28 99.3
2023-01-18 101.8 2023-01-16 99.1 2022-11-29 NA
2023-01-19 NA 2023-01-17 99.5 2022-11-30 NA

Answer 1

我会提供这个答案，直到有人想出更优雅的东西。

library(tidyverse)
dat <- structure(list(date1 = structure(c(19370, 19373, 19374, 19375, 
                                          19376), class = "Date"), Var1 = c(100.1, 104.5, 101.6, 101.8, 
                                                                            NA), date2 = structure(c(19368, 19369, 19370, 19373, 19374
                                                                            ), class = "Date"), Var2 = c(99.7, 99.8, 99.9, NA, NA), date3 = structure(c(19320, 
                                                                                                                                                            19321, 19324, 19325, 19326), class = "Date"), Var3 = c(102.3, 
                                                                                                                                                                                                                   99.9, 99.3, 100.5, 100.1)), row.names = c(NA, -5L), class = c("tbl_df", 
                                                                                                                                                                                                                                                                                 "tbl", "data.frame"))
dat2 <- dat %>%
  pivot_longer(cols = contains("date"),
               names_to = "date") %>%
  select(date, value, contains("Var")) %>%
  arrange(date) %>%
  mutate(id = group_indices(group_by(., date))) %>%
  select(contains("Var"), date = value, id)

var_nms <- names(select(dat2, contains("Var"))) 
for (i in seq_along(var_nms)){
  dat2[[var_nms[i]]] <- if_else(dat2$id == i, dat2[[var_nms[i]]],
                                NA_real_)
}

out <- dat2 %>%
  mutate(Var = do.call(coalesce, pick(contains("Var")))) %>%
  select(date, Var)

out
#> # A tibble: 15 x 2
#>    date         Var
#>    <date>     <dbl>
#>  1 2023-01-13 100. 
#>  2 2023-01-16 104. 
#>  3 2023-01-17 102. 
#>  4 2023-01-18 102. 
#>  5 2023-01-19  NA  
#>  6 2023-01-11  99.7
#>  7 2023-01-12  99.8
#>  8 2023-01-13  99.9
#>  9 2023-01-16  NA  
#> 10 2023-01-17  NA  
#> 11 2022-11-24 102. 
#> 12 2022-11-25  99.9
#> 13 2022-11-28  99.3
#> 14 2022-11-29 100. 
#> 15 2022-11-30 100.

^{创建于 2023-04-12 与 reprex v2.0.2}

Answer 2

使用末尾注释中可重复显示的数据，假设需要的是一个动物园对象，每个非日期列都有一个单独的列。

首先创建一个看起来像

g

的分组向量

c("Var1", "Var1", "Var2", "Var2", "Var3", "Var3")

，然后将

DF

转换为列表，并用

g

拆分它，得到

s

。最后将

s

的每个组件转换为动物园对象，并使用

cbind

合并它们。 （如果需要在结果上使用

fortify.zoo

的数据框。）

library(zoo)

g <- rep(names(DF)[sapply(DF, is.numeric)], each = 2)
s <- split(as.list(DF), g)
do.call("cbind", lapply(s, function(x) read.zoo(as.data.frame(x))))

给予：

            Var1 Var2  Var3
2022-11-24    NA   NA 102.3
2022-11-25    NA   NA  99.9
2022-11-28    NA   NA  99.3
2022-11-29    NA   NA    NA
2022-11-30    NA   NA    NA
2023-01-11    NA 99.7    NA
2023-01-12    NA   NA    NA
2023-01-13 100.1 99.9    NA
2023-01-16 104.5 99.1    NA
2023-01-17 101.6 99.5    NA
2023-01-18 101.8   NA    NA
2023-01-19    NA   NA    NA

注意

Lines <- "date1 Var1 date2 Var2 date3 Var3
2023-01-13 100.1 2023-01-11 99.7 2022-11-24 102.3
2023-01-16 104.5 2023-01-12 NA 2022-11-25 99.9
2023-01-17 101.6 2023-01-13 99.9 2022-11-28 99.3
2023-01-18 101.8 2023-01-16 99.1 2022-11-29 NA
2023-01-19 NA 2023-01-17 99.5 2022-11-30 NA"
DF <- read.table(text = Lines, header = TRUE)