Identify only the first matching record - r

Question

I have a large amount of time series data stored in a dataframe called "Tag.data" where one record is taken every 30 seconds over the course of several months. For example:
2013-09-30 23:59:00
2013-09-30 23:59:30
2013-10-01 00:00:00
2013-10-01 00:00:30
2013-10-01 00:01:00
2013-10-01 00:01:30
2013-10-01 00:02:00
...
2013-10-15 05:00:00
2013-10-15 05:00:30
2013-10-15 05:01:00
2013-10-15 05:01:30
2013-10-15 05:02:00
...
This data is stored in Tag.data$dt.
Within my data I would like to identify the 1st and 15th day of each month so that these can be used on a later plot.
I was successfully able to identify the first day of each month with this code:
locs <- tapply (X=Tag.data$dt, FUN=min, INDEX=format(Tag.data$dt, '%Y%m'))
at <- Tag.data$dt %in% locs
at <- at & format(Tag.data$dt, '%m') %in% c('01', '02', '03','04', '05', '06','07', '08', '09','10', '11', '12') & format(Tag.data$dt, '%d') == '01'
Unfortunately I was less successful when I attempted to also identify the 15th day of each month with this code:
locs <- tapply (X=Tag.data$dt, FUN=min, INDEX=format(Tag.data$dt, '%Y%m'))
at <- Tag.data$dt %in% locs
at <- at & format(Tag.data$dt, '%m') %in% c('01', '02', '03','04', '05', '06','07', '08', '09','10', '11', '12') & format(Tag.data$dt, '%d') == '01'|
format(Tag.data$dt, '%m') %in% c('01', '02', '03','04', '05', '06','07', '08', '09','10', '11', '12') & format(Tag.data$dt, '%d') == '15'
While this did identify both the 1st and the 15th days of each month, for some reason it identifies only one record for the 1st day of the month but every record for the 15th day of the month (of which there are a great many). I would like to identify only the first record for both the 1st and 15th days of each month. Any help would be much appreciated.

Answer 1

Judging from your code:
locs <- tapply (X=Tag.data$dt, FUN=min, INDEX=format(Tag.data$dt, '%Y%m'))
I assume Tag.data$dt is stored as one of POSIX classes.
I would like to identify only the first record for both the 1st and 15th days of each month.
Probably slow, but this does the work.
ymd <- format(Tag.data$dt,"%Y%m%d")
index.01.15 <- !duplicated(ymd) & grepl("01$|15$", ymd)
You can use the logical vector to select the rows Tag.data[index.01.15, ]

Answer 2

Try this. It makes use of lubridate. You can select all rows where the day is either 1 or 15.
library(lubridate)
options(stringsAsFactors=FALSE)
Tag.data = structure(list(dt = c("30/09/2013 23:59", "1/10/2013 0:00", "1/10/2013 0:00",
"1/10/2013 0:01", "1/10/2013 0:01", "1/10/2013 0:02", "2/10/2013 0:04",
"15/10/2013 5:00", "15/10/2013 5:00", "15/10/2013 5:01", "15/10/2013 5:01",
"15/10/2013 5:02")), .Names = "dt", class = "data.frame", row.names = c(NA,
-12L))
Tag.data$dt = parse_date_time(Tag.data$dt, '%d/%m/%Y %H%M')
at = Tag.data[day(Tag.data$dt) %in% c(1,15), ]
This is more flexible as you can specify any day you wish to subset on. E.g replace the values in c(1,15) for any day, or month(Tag.data$dt) %in% c(<INSERT MONTH NUMBER>) to subset on month.

Answer 3

It looks like your data are already stored as dates of some sort (e.g., POSIXct). Something like this, but with even more rows?
Tag.data <- data.frame(dt=seq(ISOdate(2013,10,1), by = "30 min", length.out = 10000))
Then if you want just the first record from each 1st or 15th day, this might work:
daychars <- format(Tag.data$dt, '%d')
day1or15 <- daychars %in% c("01","15")
newday <- c(TRUE, (daychars[1:(length(daychars)-1)] != daychars[2:length(daychars)]))
format(Tag.data[day1or15 & newday,"dt"],"%m/%d/%Y %H:%M:%S")
The newday line helpfully does not require that the day begins at any particular time, but it does assume that your time series is ordered.

Answer 4

I suggest you use the excellent xts package for time series data in R.
You didn't provide reproducible data, so i made some of my own.
require(xts)
Tag.data <- xts(rnorm(1e5), order.by = Sys.time() + seq(30, 3e6, 30))
Sub-setting by day of the month is a simple one-liner.
days_1n15 <- Tag.data[.indexmday(Tag.data) %in% c(1, 15)]
This returns all records on the 1st and 15th day of any month.
Now we just need to pull out the first observations on each matching day.
firstOf <- do.call(rbind, lapply(split(days_1n15, 'days'), first))
Which contains the data you want:
R> firstOf
[,1]
2014-02-01 21:29:01 1.284222
2014-02-15 00:00:01 -1.262235
2014-03-01 00:00:01 -0.465001