[R] problem applying the same function twice

Sarah Goslee sarah.goslee at gmail.com
Tue Mar 10 22:12:03 CET 2015


Yeah, that's tiny:

> fullout <- expand.grid(site=1:669, year=1:7, sample=1:3)
> dim(fullout)
[1] 14049     3


Almost certainly the problem is that your expand.grid result doesn't
have the same column names as your actual data file, so merge() is
trying to make an enormous result. Note how when I made outgrid in the
example I named the columns.

Make sure that the names are identical!


On Tue, Mar 10, 2015 at 4:57 PM, Curtis Burkhalter
<curtisburkhalter at gmail.com> wrote:
> Sarah,
>
> I have 669 sites and each site has 7 years of data, so if I'm thinking
> correctly then there should be 4683 possible combinations of site x year.
> For each year though I need 3 sampling periods so that there is something
> like the following:
>
> site 1      year1      sample 1
> site 1      year1      sample 2
> site 1      year1      sample 3
> site 2      year1      sample 1
> site 2      year1      sample 2
> site 2      year1      sample 3.....
> site 669   year7      sample 1
> site 669   year7     sample 2
> site 669   year7     sample 3.
>
> I have my max memory allocation set to the amount of RAM (8GB) on my laptop,
> but it still 'times out' due to memory problems.
>
> On Tue, Mar 10, 2015 at 2:50 PM, Sarah Goslee <sarah.goslee at gmail.com>
> wrote:
>>
>> 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