Smoking with R's tmap

In my first post, I’ll explain how to process public health data to generate a map on R.

I’ve taken a csv file with the percentage of current smokers in Scotland’s health boards from here.

You get a file with just four columns, one with the health board code, the rate of smoking for males and females, and the total rate.

Basically, I tried a few packages and got stuck with various dependency errors. I used sf (simple features) and sp to read in and join the files. For the map, I ended up using tmap which is flexible and the syntax is nice and concise.

I used st_read to read in a shapefile for health boards, and joined the smokers file to that using sp’s merge through the health board codes.

shapeData <- st_read("SG_NHS_HealthBoards_2019.shp")
shp_hb <- sp::merge(shapeData, smokers_hb, by.x = "HBCode", by.y ="hb2019", duplicateGeoms = T)

Note the “duplicateGeoms = TRUE” here, as the data files contain non-unique matches, ie. there are many rows for one health board code in the shapefile.

st_read is good because you don’t have to transform the file afterwards like with rgdal for example - if you inspect it, e.g. summary(shapeData), you see that it’s a spatial polygons dataframe and doesn’t need converting into a dataframe.

Finally, the code to create the map starts with tm_shape and the merged file. In tm_fill(col = "Rate"), you tell it to fill the polygons with colours defined by the variable you want, in this case, total rate of smoking.

I tested a few colour palettes when I was using other packages (like RColorBrewer), and when including the palette you have to define whether the data is continuous or discrete. I was confused at many points with this, as the smoking data is… not continuous but not discrete? The csv just includes a number (the percentage) for each health board (it’s discrete). But tmap has grouped the data automatically, showing the range of values and the colour they correspond to in the legend bar. This also helps to see which health boards are similar in their rates.

With other packages, I also had issues trying to centre the health board labels and not having them overlap, but tmap has a handy argument, remove_overlap which does just that. And I ended up using the tmap style cobalt.

tm_shape(shp_hb) +
  tm_fill(col = "Rate") +
  tm_text("HBName", size = 0.5, col = "white", remove.overlap = TRUE) + 
  tm_borders() +
  tmap_style("cobalt") +
  tm_layout(legend.position = c("left", "top"), title= '% current smokers in Health Boards', 
            title.position = c("left", "top")) +
  tm_credits("Scottish Health Survey, 2019", position=c("left", "bottom"))

This is the resulting map:

There are things that could be improved, like some of the health board names could be showing better, but overall I’m pleased with how it turned out.