re-arrange an unpleasant real work data frame for time series analysis in r

Question

Here I have an example of the ideal input data for time series analysis:

However I receive the raw data like this:

raw_data <- data.frame(matrix(nrow=4, ncol=5))
colnames(raw_data) <- c("site","date","00:00","01:00","02:00")
raw_data$site <- c("A","B","A","B")
raw_data$date <- c("2015-01-01","2015-01-01","2015-01-02","2015-01-02")
raw_data$`00:00` <- c(1,4,1,4)
raw_data$`01:00` <- c(2,5,2,5)
raw_data$`02:00` <- c(3,6,3,6)

I have spent really a lot of time trying to re-arrange the raw data into the ideal structure. Really appreciate any help. Thanks.

Answer 1

We can use pivot_longer to reshape to 'long' format and then with unite join the columns

library(dplyr)
library(tidyr)
library(lubridate)
raw_data %>%
  pivot_longer(cols = matches('^[0-9]'), names_to = 'Time') %>% 
  unite(DateTime, date, Time, sep=" ") %>%
  mutate(DateTime = ymd_hm(DateTime))
# A tibble: 12 x 3
#   site  DateTime            value
#   <chr> <dttm>              <dbl>
# 1 A     2015-01-01 00:00:00     1
# 2 A     2015-01-01 01:00:00     2
# 3 A     2015-01-01 02:00:00     3
# 4 B     2015-01-01 00:00:00     4
# 5 B     2015-01-01 01:00:00     5
# 6 B     2015-01-01 02:00:00     6
# 7 A     2015-01-02 00:00:00     1
# 8 A     2015-01-02 01:00:00     2
# 9 A     2015-01-02 02:00:00     3
#10 B     2015-01-02 00:00:00     4
#11 B     2015-01-02 01:00:00     5
#12 B     2015-01-02 02:00:00     6

Answer 2

You can do this using melt from the data.table package:

library(data.table)

# Mark the data as a data.table
setDT(raw_data)

# Melt it into long format
new_data <- melt(raw_data, id.vars=c('site', 'date'), variable.name='time')

# Put date and time together into a new column, and delete the old ones
new_data[, `:=`(DateTime = paste(date, time),
                date = NULL, time = NULL)]