Question

I have a dataframe:

> df <- data.frame(
+   Species = rep(LETTERS[1:4], times=c(5,6,7,6)),
+   Length = rep(11:14, each=3) 
+ )
> 
> df

I need to be able to count the number of individuals of a certain Length for each Species (i.e., how many individuals in Species A have a length of 1, 2, 3, etc?) Then, I need to perform a series of additional analyses on the output. For example, I need to calculate the density of individuals of each length, and the decrease in density from one length class to the next.

This is easy if I subset the data first:

Spec.A<-df[df$Species=="A",] 

#count number of specimens of each length; 
count<-table(Spec.A$Length)
count

#calculate density per length category (divide by total area sampled =30) 
density<-count/(30)
density

#calculate the decrease in density (delta.N) from one length category to the next; 
delta.N<-diff(density, lag=1, differences=1)
delta.N

The problem is that I need to do these calculations for each species (i.e., to loop through each subset).

On the one hand, I could use tapply(), with a function that uses table();

#function: count number of specimens of each length; 
count<-function(x){
table(x)
}

Number<-tapply(df$Length, df$Species, FUN=count, simplify=FALSE)
Number

This gives me what I want, but the format of the output is funky, and I can’t figure out how to perform additional analyses on the results.

I have tried using ddply() from plyr, something like:

ddply(df$Length, df$Species,
count)

But I clearly don’t have it right, and I’m not even sure ddply() is appropriate for my problem, given that I have a different number of length observations for each species.

Should I be looking more closely at other options in plyr? Or is there a way to write a for loop to do what I need?

Answer 1

You’re on the right track! tapply with list output is definitely one way to go, and may be a good choice since your outputs will have varying lengths.

ddply, like you guessed, is another way. The key is that the output of the function you give to ddply should be a data frame with all your statistics in a “long” mode (so that they will stack nicely). The simple count function can’t do this, so you’ll need to make your own function. The way I go about devising a function for a ddply call like this is actually very similar to what you were doing: I get a subset of the data, and then craft my function using that. Then, when you submit it to ddply, it’ll nicely apply that function across all the subsets.

SpeciesStats <- function(df) {
  counts    = table(df$Length)
  densities = counts/30
  delta.N   = diff(densities, lag=1, differences=1)

  data.frame(Length   = names(counts),
             Count    = as.numeric(counts),
             Density  = as.numeric(densities),
             delta.N  = c(NA, delta.N), 
             row.names=NULL)
}

> ddply(df, 'Species', SpeciesStats)
   Species Length Count    Density     delta.N
1        A     11     3 0.10000000          NA
2        A     12     2 0.06666667 -0.03333333
3        B     12     1 0.03333333          NA
4        B     13     3 0.10000000  0.06666667
5        B     14     2 0.06666667 -0.03333333
6        C     11     3 0.10000000          NA
7        C     12     3 0.10000000  0.00000000
8        C     14     1 0.03333333 -0.06666667
9        D     13     3 0.10000000          NA
10       D     14     3 0.10000000  0.00000000