Question

I am trying to run a jack-knife using Plyr. I have a large dataset (715 sites over 10 years). I have already calculated the Species Richness (count of all species present) in a square for each year but now I want to calculate new Richness values, after taking out one species at a time and have them all in one dataset.

Example data:

Site <- c(1,1,1,1,1,1)
Year <- c(96,96,96,97,97,97)
SpID <- c(1,2,3,1,2,3)
Count <- c(1,1,1,1,1,1)
data <- cbind(Site, Year, SpID)

So overall for Site 1 the species richness is 3 in both years. If I want recalculate this without one of the species it would now be 2.

I have tried using the following code:

foo<-function(z){
    data2 <- subset(data, SpID != (z))
    summaryBy(Count~ Year + Site, 
              data = data2, 
              FUN = function(x) { c(l = length(x)) } )
}

richall<- ddply(data,.(SpID),foo)

But I’m obviously making a mistake somewhere! Any thoughts?

Answer 1

With your example data and call to ddply, this is what will happen:

• ddply will find the different values in the SpID column of your
  dataset (1, 2 and 3)
• It will next create a data.frame for each of these unique values.
• Each of these data.frames will hold only the rows for which the
  SpID is equal to that unique value (so: a data.frame with the first
  and fourth rows, one with the second and fifth and one with the third
  and last rows)
• Function foo will now be called, passing each of these data.frames
  one at a time as its first argument

So it is rather obvious now that this will not help in doing jack-knife. In fact I don’t see an obvious way of attaining that with plyr. In this particular case you’re probably better off rigging your own with similar logic. Something like:

listOfResults <- 
    lapply(unique(data$SpID), 
           function(curID) {
               curDF<-data[data^SpID!=curID,]
               summaryBy(...,data=curDF)
           })

You can then recombine your results with e.g. ?do.call.