[R] problem applying the same function twice

[Curtis Burkhalter]

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

https://sites.google.com/site/curtisburkhalter/

	[[alternative HTML version deleted]]