R: Kernel Density Plots (Bandwidth Must be Strictly Positive)

Question

I am using the R programming language. I am following this tutorial over here for making 3d kernel density plots in R: https://plotly.com/r/3d-surface-plots/ :

library(MASS)
library(plotly)

kd <- with(MASS::geyser, MASS::kde2d(duration, waiting, n = 50))
fig <- plot_ly(x = kd$x, y = kd$y, z = kd$z) %>% add_surface()

fig

I decided to try this on my own data:

#generate data
a = rnorm(100,10,10)
b = rnorm(100,5,5)
c = rnorm(100,5,10)
d = data.frame(a,b,c)

#make 3d plot (I think n = 50 refers to selecting the first 50 points?)
kd <- with(d, MASS::kde2d(a,b,c, n = 50))
fig <- plot_ly(x = kd$x, y = kd$y, z = kd$z) %>% add_surface()

But this results in the following error:

Error in MASS::kde2d(a, b, c, n = 50) : 
  bandwidths must be strictly positive

This error prevents me from creating the "kd" object.

Can someone please tell me what am I doing wrong? Is there a problem with the specific data I am using? Or is this a syntax error?

Thanks

Answer 1

You seem to be misunderstanding the purpose of kde2d . From help(kde2d) :

Two-dimensional kernel density estimation with an axis-aligned bivariate normal kernel, evaluated on a square grid.

From the same help file regarding the h argument:

h
vector of bandwidths for x and y directions. Defaults to normal reference bandwidth (see bandwidth.nrd). A scalar value will be taken to apply to both directions.

You are passing c , a length 100 numeric vector as h . You appear to be trying to pass data to h , this does not make sense. Pass either one or two values for bandwidth or nothing and accept the default.

From lines 31 and 32 of the source , we can see why you got the error:

    if (any(h <= 0))
        stop("bandwidths must be strictly positive")

Thus, if either of the first two values of c are negative or zero, you will get this error.

---

The n argument, as described in the help file:

n
Number of grid points in each direction. Can be scalar or a length-2 integer vector.

This determines the grid that the density is provided. If you provide a single value, a square grid is produced.