R: Problem with bandwith selection of kernel density

Question

I want to calculate the optimal bandwith value for my kernel density estimation. I have a .csv with two columns (longitude and latitude). I tried several different functions but always get different errors. So far i tried:

h.amise(x, deriv.order = 0)

which give my following error: argument 'x' must be numeric and need at least 3 data points However, I checked my dataframe and it is numeric.

Then I tried:

dpik(x)

which give my following error: 'list' object cannot be coerced to type 'double'

Is is wrong to use a .csv as dataframe with two columns or what can be the issue?

Answer 1

Based on your sample data:

library(kedd)
library(KernSmooth)

h.amise(x$long)
# 
# Call:     Aymptotic Mean Integrated Squared Error
# 
# Derivative order = 0
# Data: x$long (10 obs.);   Kernel: gaussian
# AMISE = 0.02761525;   Bandwidth 'h' = 1.57264

h.amise(x$lat)
# 
# Call:     Aymptotic Mean Integrated Squared Error
# 
# Derivative order = 0
# Data: x$lat (10 obs.);    Kernel: gaussian
# AMISE = 0.01352194;   Bandwidth 'h' = 3.37266

dpik(x$long)
# [1] 0.4912055
dpik(x$lat)
# [1] 1.079109

Read the manual pages ?h.amise and ?dpik for the details. The functions take a single numeric vector as input.