In ellieallien/Setviz: Visualise intersecting sets with weights

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(Setviz)

Example data

Start with a dataset with logical or binary 0/1 columns:

data[,c("WASH","Protection","Education","Shelter","Livelihoods")] %>% head

Basic Plot

You can create the intersection plot directly:

plot_set_percentages(data = data,
                     varnames = c("WASH","Protection","Education","Shelter","Livelihoods"),
                     nintersects = 12)

Cummulative vs. Exclusive sets

By default, the sets are calcualted cumulatively. This means that in this example in this example, the set "Wash & Shelter" also includes all records that are in "Wash & Shelter & Livelihoods".

To prevent this behaviour and have exact, mutually exclusive sets (meaning that "Wash & Shelter" contains exclusively the records that are TRUE for Wash and shelter and FALSE for all other variables), you can set the mutually_exlusive_sets parameter to TRUE:

plot_set_percentages(data = data,
                             varnames = c("WASH","Protection","Education","Shelter","Livelihoods"),nintersects = 12,
                             mutually_exclusive_sets = TRUE)

Include single category sets

To include sets of a single category (no intersection but only one of the binary data variables are 1/TRUE), set exclude_unique to FALSE.

plot_set_percentages(data = data,
                             varnames = c("WASH","Protection","Education","Shelter","Livelihoods"),nintersects = 12,
                             exclude_unique = FALSE)

As you see above, if you are using cumulative sets, the single-category sets will always be larger than combinations that contain them. This option makes most sense with mutually exclusive sets:

plot_set_percentages(data = data,
                             varnames = c("WASH","Protection","Education","Shelter","Livelihoods"),nintersects = 12,
                             mutually_exclusive_sets = TRUE,
                             exclude_unique = FALSE)

IMPORTANT NOTES

For complex samples that require weighted percentages, you need to provide the weight_variable or the weight_function paramter. (please see ?plot_set_percentages for details.)
The "Set Size" on the bottom left may not be correct and should not be included in any published outputs.
Currently decimal percentages are not supported. Therefore sets with <=0.5% share will be ignored. You can instead treat them as 1% by setting round_to_1_percent to TRUE