Format the 'to'-break values differently

[mcSamuelDataSci]

This issue may be related to the issue you already solved that I submitted #249. And, I may be doing some weird stuff, and am admittedly mapping a very awkward and skewed "distribution", BUT, the plot mode map is more-or-less clear and certainly interpretable.

The same map in view mode has some important problem with the legend, such that a color in the map does not appear in the legend, so it become uninterpretable.

Here is some code that reproduces the exact problem, and has some comments to show what I am doing.

# load needed packages for the current R session
library(dplyr)
library(tigris)   
library(tmap)
library(classInt)  
library(RColorBrewer)

# Create a California census tract object
options(tigris_class = "sf")  
caTract  <- tracts(state = "CA", cb = TRUE)  

# Read .csv file from GitHub that links tracts to "communities" (MSSAs)
link_tract_community  <- read.csv("https://raw.githubusercontent.com/mcSamuelDataSci/CACommunityBurden/master/myCBD/myInfo/cbdLinkCA.csv",colClasses = "character")

# Merge/join census tract shape file with tract-community link file
caTract <- caTract %>% 
  left_join(link_tract_community, by="GEOID")

# Create a new "shape file" by grouping the tracts by communities. 
shape_Comm  <- caTract %>% 
  group_by(county,comID) %>% 
  summarize()  %>% ungroup()

# Read ("fake") bit of California death data
githubURL <- "https://raw.githubusercontent.com/mcSamuelDataSci/CACommunityBurden/master/myCBD/myData/fake/conditionX.RDS"
download.file(githubURL,"temp.rds", method="curl")
community_data <- readRDS("temp.rds")

# Merge shape file object to the community death data object.
map_work <-  left_join(shape_Comm,community_data, by=c("county","comID"))

map_work$Ndeaths[is.na(map_work$Ndeaths)] <- 0

# Out of context, this line below might seem bad/weird (i.e. why not put in the equivalent in the tm_polygons below),
#    but it is actually important for my project
myBreaks  <- classIntervals(map_work$Ndeaths,style="quantile",breaks=NULL,n=5)$brks
myPal     <- rev(brewer.pal(5,"RdYlBu"))

xmap <- tm_shape(map_work) + tm_polygons(col="Ndeaths",palette = myPal, style="fixed",breaks=myBreaks, 
                              interval.closure="right",
                              legend.reverse=T
                              )

# this looks "fine".  Yes, it is "bad" that there are three colors for "0", but at least it is
#   interpretable
tmap_mode("plot")
xmap

# this is where something is wrong:
#  "legend.reverse=T" is ignored
#  there is orange in the map, but no orange in the legend
#  There are 3 levels in the legend, not 5
tmap_mode("view")
xmap

# yes, the distribution of this variable being mapped is very skewed (sort of by definition, since
# all values less that 11 are supressed and shown as 0)
table(map_work$Ndeaths)

I sure appreciate any help you can provide!

[mcSamuelDataSci]

Hi Martijn-
I know this is a odd little issue that might only arise with a small, highly skewed, distribution being mapped. Nevertheless, if there is any "proper" fix that could implement for this, that would be most appreciated. If not, I can do some "work arounds" on my end.
Thanks! Michael

[mtennekes]

Hi Michael,

This was a common problem: leaflet::addLegend has its own interpretation of which items are included.
I did solve the problem by calling addLegend manually. So it should work now!

Could you do me a favor? Could you test the legend behavior in interactive mode, also for other data/map types?

Thx!
Martijn

[mcSamuelDataSci]

Martijn-
Works great based on a bit of testing! Thank you so much. I will add an acknowledgement of your assistant on our project site, unless you would prefer I don't for any reason?

Also, I will dig into it more myself, but can you direct me to anything I might look at such that a legend might show something along the lines of 0-4.9, 5.0-9.9 etc. (rather than 0-5, 5-10, etc) even if the data is actually continuous? Perhaps no simple way.

Thanks again, Michael

[mtennekes]

You're welcome, Michael!

There is no easy way to format the numbers like that. There is an argument legend.format, which can be used to format the legend numbers, also with a custom function. However, this function does not distinguish 'from' and 'to' numbers.

I can image applications for which this could also be handy, say for an integer variable from 1 to 100, which is considered as continuous variable, and where the preferred breaks are 1-9, 10-19, 20-29, etc., rather than 1-10, 10-20, 20-30, etc.

[Nowosad]

See new improvements in this area at https://github.com/mtennekes/tmap/issues/406#issuecomment-610559092.