[R] problem applying the same function twice

[Sarah Goslee]

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