Trying to generate and store multiple KDE models in R

Question

I am working with a point dataset that includes data from several gps collars on animals with unique identification numbers.

What I am trying to do is model their ranges using Kernel Density Estimator for each animal. What I have produces a graph of each model but I need to save the actual estimator for each one for further analyses. I'm going to reveal my ignorance here, but I tried to do this with a for loop.

collars<-unique(gps$CollarID)

where gps is my dataframe

for (i in 1:length(collars)){
    collar<-subset(gps, CollarID == collars[i], select=c(Easting, Northing))
    cxy<-cbind(collar$Easting,collar$Northing)
    kde<-kde(cxy)
    plot(kde, xlab = "X", ylab = "Y")
    title(main = collars[i])
}

What I am after is producing a uniquely named kde object for each itteration. I have tried to include the counter in the name of the object, but quickly found that that did not work.

Any help would be greatly appreciated!

-JF

Answer 1

Using dplyr, you could wrap the calculations in a function that calculates kde for each Easting, Northing observation and save resultant kde objects in list to be plotted by a separate function

Since we do not have the sample dataset, I have used mtcars dataset.

#load libraries, assuming you are using kde function from "ks" library
library("dplyr")
library("lazyeval") #required for interp function
library("ks")



DF = mtcars
indexColName = "cyl"
indexVar = unique(DF[,indexColName])
calcVars = c("hp","wt")


### Replacement values for your dataset ###

# DF = gps
# indexColName = "CollarID"
# indexVar = unique(DF[,indexColName])
# calcVars = c("Easting","Northing")

kde calc function:

fn_kdeCalc <- function(indexVarInput = indexVar,indexColInput = indexColName,calcVarsInput=calcVars) {

cat("Begin kde calc for",indexColInput,"=",indexVarInput,"\n")

#filter condition logic translates as CollarID == collars[i]
#the following is required for dynamic inputs , you can read more from trying, vignette("nse")

filter_criteria <- interp(~ filter_column == indexVarInput, filter_column = as.name(indexColInput))

kdeObj <- DF %>% 
        dplyr::filter_(filter_criteria) %>%           # subset dataset to specific filter condition
        dplyr::select(one_of(calcVarsInput)) %>%      # select specific columns for kde calc
        kde()                                         # calculate kde for selected variables

cat("End kde calc for",indexColInput,"=",indexVarInput,"\n")

return(kdeObj)

}

#calculate kde for each unique input variable using above function and save in a list

kdeObjList = lapply(indexVar,function(x) { 

fn_kdeCalc(indexVarInput = x,indexColInput = indexColName,calcVarsInput=calcVars) 

})

kde plot function:

fn_kdePlot = function(kdeObject = NULL,titleObj = NULL,labelVars=calcVars) {
  plot(kdeObject, xlab = labelVars[1], ylab = labelVars[2],main = paste0("Kernel Density estimation for ",indexColName,":",titleObj) )
}



### save plots ###
# you can use png() to save in different file for each plot or 
# pdf("KDE_plots1.pdf") to include them in one file

png()

lapply(1:length(kdeObjList), function(x) fn_kdePlot(kdeObject = kdeObjList[[x]],titleObj = indexVar[x]) )

dev.off()