如何将按因子级别划分的列表重新组合到原始数据框？

Question

I have some tracking data where I want to calculate the time difference between each point which I can do with this:

# prep the data
ID = c(rep("A",5), rep("B",5))
DateTime = c("2014-09-25 08:39:45", "2014-09-25 08:39:48", "2014-09-25 08:40:44", "2014-09-25 09:04:00","2014-09-25 09:04:10", "2014-09-25 08:33:32", "2014-09-25 08:34:41", "2014-09-25 08:35:24", "2014-09-25 09:04:00", "2014-09-25 09:04:09")
speed = c(1:10)
df = data.frame(ID,DateTime,speed, stringsAsFactors = FALSE)
df$DateTime<-as.POSIXct(df$DateTime, tz = "UTC")

# function to calculate time differences 
timeCheck<-function(df) {
  sapply(1:(nrow(df) - 1), function(i){
    timeDiff<- difftime(df$DateTime[i+1], df$DateTime[i], units = "sec" )
    return(timeDiff)
  })
}
# preserve order of factor levels 
df$ID <- factor(df$ID, levels=unique(df$ID))

# apply the function by ID
timeDiffData<-sapply(split(df, df$ID), timeCheck)

I want to be able to add a new column of the time differences to the original dataframe but of course this list is a different length, because the function doesn't calculate the time difference from itself.

I then want to use these time differences in a new function to split the tracks if the difference is greater than a certain value (say 100 seconds for the sake of example) and have the ID reflect this.

So in the end I'd have 4 levels for my ID column and the split would occur when the time difference is > 100 seconds.

The resulting dataframe should look something like:

# what it should look like 
ID = c(rep("A",3),rep("A1",2) , rep("B",3), rep("B1",2))
DateTime = c("2014-09-25 08:39:45", "2014-09-25 08:39:48", "2014-09-25 08:40:44", "2014-09-25 09:04:00","2014-09-25 09:04:10", "2014-09-25 08:33:32", "2014-09-25 08:34:41", "2014-09-25 08:35:24", "2014-09-25 09:04:00", "2014-09-25 09:04:09")
speed = c(1:10)
timeDiff<-c(NA,3,56,1396,10,NA,69,43,1716,9)
newdf = data.frame(ID,DateTime,speed,timeDiff, stringsAsFactors = FALSE)
newdf$DateTime<-as.POSIXct(df$DateTime, tz = "UTC")
newdf

Answer 1

Really your operation has three steps:

• Group your data by ID
• Compute the time differences between each timestamp in your group (the first time difference is NA)
• Create a new ID that counts the number of prior time gaps that are large (eg > 100 seconds)

This can be done pretty simply with dplyr , using group_by for the grouping and mutate for computing new variables within each group:

library(dplyr)
df %>%
  group_by(ID) %>%
  mutate(timeDiff = c(NA, difftime(tail(DateTime, -1), head(DateTime, -1), units="sec"))) %>%
  mutate(newID = paste0(ID, cumsum(!is.na(timeDiff) & timeDiff > 100))) %>%
  ungroup()
# A tibble: 10 × 5
#       ID            DateTime speed timeDiff newID
#    <chr>              <dttm> <int>    <dbl> <chr>
# 1      A 2014-09-25 08:39:45     1       NA    A0
# 2      A 2014-09-25 08:39:48     2        3    A0
# 3      A 2014-09-25 08:40:44     3       56    A0
# 4      A 2014-09-25 09:04:00     4     1396    A1
# 5      A 2014-09-25 09:04:10     5       10    A1
# 6      B 2014-09-25 08:33:32     6       NA    B0
# 7      B 2014-09-25 08:34:41     7       69    B0
# 8      B 2014-09-25 08:35:24     8       43    B0
# 9      B 2014-09-25 09:04:00     9     1716    B1
# 10     B 2014-09-25 09:04:09    10        9    B1

Answer 2

One answer that worked perfectly was deleted by the author. Here it is for posterity:

library(data.table)
setDT(df)[ , ID2 := paste0(ID, cumsum(c(0, diff(DateTime)) > 100)), by = ID]