Using the last day in each month of my time series in R-CodePudding

I need to use only the last day available in my dataset to aggregate later on but I didn´t have success...

library(tibbletime)
      
dataset <- data.frame(
  timestamp = c("2010-01-01", "2010-01-03", "2010-01-23")
  var =       c( 1,             4,            11)
)

monthly_dataset <- as_tbl_time(dataset, index = timestamp) %>%
                   as_period("1 month")

How can I use some function or R package to aggregate my dataset only for using the last day avaiable ?

CodePudding user response：

An option could be the lubridate package, e.g.

 library(lubridate)
 library(dplyr)
    dataset <- data.frame(
      timestamp = c("2010-01-01", "2010-01-03",
     "2010-01-23", "2010-02-01", "2010-02-03", "2010-02-23"),
      var = c(1, 4, 11, 1, 4, 11)
    )
    
    
    dataset %>%
      mutate(month = timestamp %>% ymd() %>% month()) %>%
      group_by(month) %>%
      slice_tail()

Outcome:

# A tibble: 2 x 3
# Groups:   month [2]
  timestamp    var month
  <chr>      <dbl> <dbl>
1 2010-01-23    11     1
2 2010-02-23    11     2

CodePudding user response：

The answer from Julian is a nice start, but it won't work across multiple years because the grouping variable doesn't include information about the year.

The typical way to do this is to group on year-month, and then filter to the max date per year-month group.

Also, as the creator of tibbletime I would highly suggest that you no longer use it. It is deprecated and is no longer being supported. You should just use clock/lubridate for date handling alongside the tidyverse packages like dplyr, or you should use tsibble if you really need to go all in on time series.

library(lubridate)
library(dplyr)

dataset <- tibble(
  timestamp = c(
    "2010-01-01", "2010-01-03", "2010-01-23", 
    "2010-02-01", "2010-02-03", "2011-02-23"
  ),
  var = c(1, 4, 11, 1, 4, 11)
)
dataset <- mutate(dataset, timestamp = ymd(timestamp))

dataset <- dataset %>%
  mutate(
    year_month = floor_date(timestamp, "month"),
    day = day(timestamp)
  )

dataset %>%
  group_by(year_month) %>%
  filter(day == max(day)) %>%
  ungroup()
#> # A tibble: 3 × 4
#>   timestamp    var year_month   day
#>   <date>     <dbl> <date>     <int>
#> 1 2010-01-23    11 2010-01-01    23
#> 2 2010-02-03     4 2010-02-01     3
#> 3 2011-02-23    11 2011-02-01    23

^{Created on 2022-05-18 by the reprex package (v2.0.1)}