[R] problem applying the same function twice
Sarah Goslee
sarah.goslee at gmail.com
Tue Mar 10 21:50:13 CET 2015
You said your data only had 14000 rows, which really isn't many.
How many possible combinations do you have, and how many do you need to add?
On Tue, Mar 10, 2015 at 4:35 PM, Curtis Burkhalter
<curtisburkhalter at gmail.com> wrote:
> Sarah,
>
> This strategy works great for this small dataset, but when I attempt your
> method with my data set I reach the maximum allowable memory allocation and
> the operation just stalls and then stops completely before it is finished.
> Do you know of a way around this?
>
> Thanks
>
> On Tue, Mar 10, 2015 at 2:04 PM, Sarah Goslee <sarah.goslee at gmail.com>
> wrote:
>>
>> Hi,
>>
>> I didn't work through your code, because it looked overly complicated.
>> Here's a more general approach that does what you appear to want:
>>
>> # use dput() to provide reproducible data please!
>> comAn <- structure(list(animals = c("bird", "bird", "bird", "bird",
>> "bird",
>> "bird", "dog", "dog", "dog", "dog", "dog", "dog", "cat", "cat",
>> "cat", "cat"), animalYears = c(1L, 1L, 1L, 2L, 2L, 2L, 1L, 1L,
>> 1L, 2L, 2L, 2L, 1L, 1L, 1L, 2L), animalMass = c(29L, 48L, 36L,
>> 20L, 34L, 34L, 21L, 28L, 25L, 35L, 18L, 11L, 46L, 33L, 48L, 21L
>> )), .Names = c("animals", "animalYears", "animalMass"), class =
>> "data.frame", row.names = c("1",
>> "2", "3", "4", "5", "6", "7", "8", "9", "10", "11", "12", "13",
>> "14", "15", "16"))
>>
>>
>> # add reps to comAn
>> # assumes comAn is already sorted on animals, animalYears
>> comAn$reps <- unlist(sapply(rle(do.call("paste",
>> comAn[,1:2]))$lengths, seq_len))
>>
>> # create full set of combinations
>> outgrid <- expand.grid(animals=unique(comAn$animals),
>> animalYears=unique(comAn$animalYears), reps=unique(comAn$reps),
>> stringsAsFactors=FALSE)
>>
>> # combine with comAn
>> comAn.full <- merge(outgrid, comAn, all.x=TRUE)
>>
>> > comAn.full
>> animals animalYears reps animalMass
>> 1 bird 1 1 29
>> 2 bird 1 2 48
>> 3 bird 1 3 36
>> 4 bird 2 1 20
>> 5 bird 2 2 34
>> 6 bird 2 3 34
>> 7 cat 1 1 46
>> 8 cat 1 2 33
>> 9 cat 1 3 48
>> 10 cat 2 1 21
>> 11 cat 2 2 NA
>> 12 cat 2 3 NA
>> 13 dog 1 1 21
>> 14 dog 1 2 28
>> 15 dog 1 3 25
>> 16 dog 2 1 35
>> 17 dog 2 2 18
>> 18 dog 2 3 11
>> >
>>
>> On Tue, Mar 10, 2015 at 3:43 PM, Curtis Burkhalter
>> <curtisburkhalter at gmail.com> wrote:
>> > Hey everyone,
>> >
>> > I've written a function that adds NAs to a dataframe where data is
>> > missing
>> > and it seems to work great if I only need to run it once, but if I run
>> > it
>> > two times in a row I run into problems. I've created a workable example
>> > to
>> > explain what I mean and why I would do this.
>> >
>> > In my dataframe there are areas where I need to add two rows of NAs (b/c
>> > I
>> > need to have 3 animal x year combos and for cat in year 2 I only have
>> > one)
>> > so I thought that I'd just run my code twice using the function in the
>> > code
>> > below. Everything works great when I run it the first time, but when I
>> > run
>> > it again it says that the value returned to the list 'x' is of length 0.
>> > I
>> > don't understand why the function works the first time around and adds
>> > an
>> > NA to the 'animalMass' column, but won't do it again. I've used
>> > (print(str(dataframe)) to see if there is a change in class or type when
>> > the function runs through the original dataframe and there is for
>> > 'animalYears', but I just convert it back before rerunning the function
>> > for
>> > second time.
>> >
>> > Any thoughts on this would be greatly appreciated b/c my actual data
>> > dataframe I have to input into WinBUGS is 14000x12, so it's not a
>> > trivial
>> > thing to just add in an NA here or there.
>> >
>> >>comAn
>> > animals animalYears animalMass
>> > 1 bird 1 29
>> > 2 bird 1 48
>> > 3 bird 1 36
>> > 4 bird 2 20
>> > 5 bird 2 34
>> > 6 bird 2 34
>> > 7 dog 1 21
>> > 8 dog 1 28
>> > 9 dog 1 25
>> > 10 dog 2 35
>> > 11 dog 2 18
>> > 12 dog 2 11
>> > 13 cat 1 46
>> > 14 cat 1 33
>> > 15 cat 1 48
>> > 16 cat 2 21
>> >
>> > So every animal has 3 measurements per year, except for the cat in year
>> > two
>> > which has only 1. I run the code below and get:
>> >
>> > #combs defines the different combinations of
>> > #animals and animalYears
>> > combs<-paste(comAn$animals,comAn$animalYears,sep=':')
>> > #counts defines how long the different combinations are
>> > counts<-ave(1:nrow(comAn),combs,FUN=length)
>> > #missing defines the combs that have length less than one and puts it in
>> > #the data frame missing
>> > missing<-data.frame(vals=combs[counts<2],count=counts[counts<2])
>> >
>> > genRows<-function(dat){
>> > vals<-strsplit(dat[1],':')[[1]]
>> > #not sure why dat[2] is being converted to a string
>> > newRows<-2-as.numeric(dat[2])
>> > newDf<-data.frame(animals=rep(vals[1],newRows),
>> > animalYears=rep(vals[2],newRows),
>> > animalMass=rep(NA,newRows))
>> > return(newDf)
>> > }
>> >
>> >
>> > x<-apply(missing,1,genRows)
>> > comAn=rbind(comAn,
>> > do.call(rbind,x))
>> >
>> >> comAn
>> > animals animalYears animalMass
>> > 1 bird 1 29
>> > 2 bird 1 48
>> > 3 bird 1 36
>> > 4 bird 2 20
>> > 5 bird 2 34
>> > 6 bird 2 34
>> > 7 dog 1 21
>> > 8 dog 1 28
>> > 9 dog 1 25
>> > 10 dog 2 35
>> > 11 dog 2 18
>> > 12 dog 2 11
>> > 13 cat 1 46
>> > 14 cat 1 33
>> > 15 cat 1 48
>> > 16 cat 2 21
>> > 17 cat 2 <NA>
>> >
>> > So far so good, but then I adjust the code so that it reads (**notice
>> > the
>> > change in the specification in 'missing' to counts<3**):
>> >
>> > #combs defines the different combinations of
>> > #animals and animalYears
>> > combs<-paste(comAn$animals,comAn$animalYears,sep=':')
>> > #counts defines how long the different combinations are
>> > counts<-ave(1:nrow(comAn),combs,FUN=length)
>> > #missing defines the combs that have length less than one and puts it in
>> > #the data frame missing
>> > missing<-data.frame(vals=combs[counts<3],count=counts[counts<3])
>> >
>> > genRows<-function(dat){
>> > vals<-strsplit(dat[1],':')[[1]]
>> > #not sure why dat[2] is being converted to a string
>> > newRows<-2-as.numeric(dat[2])
>> > newDf<-data.frame(animals=rep(vals[1],newRows),
>> > animalYears=rep(vals[2],newRows),
>> > animalMass=rep(NA,newRows))
>> > return(newDf)
>> > }
>> >
>> >
>> > x<-apply(missing,1,genRows)
>> > comAn=rbind(comAn,
>> > do.call(rbind,x))
>> >
>> > The result for 'x' then reads:
>> >
>> >> x
>> > [[1]]
>> > [1] animals animalYears animalMass
>> > <0 rows> (or 0-length row.names)
>> >
>> > Any thoughts on why it might be doing this instead of adding an
>> > additional
>> > row to get the result:
>> >
>> >> comAn
>> > animals animalYears animalMass
>> > 1 bird 1 29
>> > 2 bird 1 48
>> > 3 bird 1 36
>> > 4 bird 2 20
>> > 5 bird 2 34
>> > 6 bird 2 34
>> > 7 dog 1 21
>> > 8 dog 1 28
>> > 9 dog 1 25
>> > 10 dog 2 35
>> > 11 dog 2 18
>> > 12 dog 2 11
>> > 13 cat 1 46
>> > 14 cat 1 33
>> > 15 cat 1 48
>> > 16 cat 2 21
>> > 17 cat 2 <NA>
>> > 18 cat 2 <NA>
>> >
>> > Thanks
>> > --
>> > Curtis Burkhalter
>
>
More information about the R-help
mailing list