1

I'm working with a two-column data.frame with the columns containing x and y coordinates (Longitude, Latitude). A simple way to visualize a density map is by using the 'stat_density_2d' function. When I run the code below on my data, I obtain large density values. To me, this density plot suggest that the densities are larger than the total number of observations, which does not make sense to me. I have provided a small subset of my data.

df <- structure(list(Lat = c(-24.1871741, -24.2069615, -24.2022726, 
    -24.2016188, -24.2152107, -24.1939073, -24.1913561, -24.198409, 
    -24.2088875, -24.2121186), Long = c(30.8839167, 30.8814249, 30.8788437, 
    30.8903969, 30.8883906, 30.8784664, 30.870561, 30.8800543, 30.8818679, 
    30.8914805)), row.names = c(NA, 10L), class = "data.frame")

ggplot(df, aes(Long, Lat)) +
    stat_density2d(geom="tile", aes(fill = ..density..), contour = FALSE) + 
    geom_point(colour = "white")

enter image description here

Even with only ten data points, the plot suggests a density range between 500 and 2000 points. I don't believe the output is wrong, but I'm just wondering if some could explain to me why these values are so larger on such a small dataset. How should I interpreted the density values from the legend?

1
  • I would highly recommend projecting your data. Under the hood ggplot2::stat_density2d is calling MASS::kde2d for the density estimate and MASS:bandwidth.nrd as the automatic bandwidth plugin. I cannot imagine that the automatic bandwidth selector is coming up with anything relevant using lat/long coordinates. Geographic coordinates are degrees not distance. Jul 8 '20 at 16:19
0

density is points per unit area and your area is about

(30.890-30.870)*(24.215-24.190)

0.0005 square degrees. There's 10 points, which means that's about 20,000 points per unit area. The smoothing done by the density estimation brings that down.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.