Question

I have a data.frame that looks like this:

ID     Date.A        Date.B        Variable
A      01/01/2012    03/24/2012    Apples
A      02/01/2012    03/24/2012    Oranges
B      01/01/2012    02/04/2012    Bananas
C      01/01/2012    04/22/2012    Apples
A      03/01/2012    03/24/2012    Grapes
B      02/01/2012    02/04/2012    Oranges
D      01/01/2012    03/15/2012    Apples

For each ID, I need to find the Variable which corresponds to the first of the month previous to Date.B. So for ID A, I need to extract “Oranges”, because 03/24/2012 – 1 month = 02/24/2012, and the first of that month is 02/01/2012.

I have no problem with the date calculations using lubridate: format.Date(df$Date.B – months(1), format=”%Y-%m-01″). I’m having trouble incorporating that logic into ddply or in a programmatic manner without writing a nasty for loop. Not asking for the algorithmic help or for someone to solve it for me, but would love to see an example of ddply that uses custom date logic similar to this.

Thanks in advance.

Answer 1

This is not with lubridate or plyr and I’m not even sure if it’s the solution to your problem. I chose to use the mondate package for the reasons tcash21 states:

10/31/2012 - months(1)

And I don’t use plyr much any more so I don’t really think in terms of plyr that much. if your data is called dat:

#Read your data in
dat <- read.table(text="ID     Date.A        Date.B        Variable
A      01/01/2012    03/24/2012    Apples
A      02/01/2012    03/24/2012    Oranges
B      01/01/2012    02/04/2012    Bananas
C      01/01/2012    04/22/2012    Apples
A      03/01/2012    03/24/2012    Grapes
B      02/01/2012    02/04/2012    Oranges
D      01/01/2012    03/15/2012    Apples", header=TRUE)

#This may be unnecessary but when I read your data in the dates were characters
lapply(2:3,  function(i) {dat[, i] <<- as.Date(dat[, i], "%m/%d/%Y")})



library(mondate)
prevmon <- as.character(mondate(dat$Date.B) - 1)  #get minus 1 month
z <- strsplit(prevmon, "-")                       #make it that first of that month
dat$Date.C <- as.Date(sapply(z, function(x) paste(x[1], x[2], "01", sep="-")))

key <- split(dat, dat$ID)                         #make list of data frames by ID 

#fruit finding function
fruiter <- function(x) x[match(x[, "Date.C"][1], x[, "Date.A"]), "Variable"]
sapply(key, fruiter)

Which yields:

      A       B       C       D 
Oranges Bananas    <NA>    <NA> 

Clearly there’s an NA for some months for the reason in the comment I attached to your question.