In uRos2018/categorical: A Better Type for Categorical Data

knitr::opts_chunk$set(echo = TRUE)

factor

Issues with factor:

• No match between labels and codes
• No support for hierarchical structure
• In (production) code, codes (at different levels) are often used
• For visualisation/tabulation/analysis labels are often more practical

Result

Current result of unConf:

• 'classification' object
• Allow user to 'cast' codes to
  • labels
  • higher levels

Example classification: business actvity (NACE)

library(dplyr)
library(categorical)
data("nace_rev2")
print(nace_rev2, short = TRUE)

Example

df <- data.frame(
  id = 1:200,
  activity = sample(nace_rev2[[4]]$id, 200),
  some_number = runif(200),
  stringsAsFactors = FALSE
)

print(head(df))

table(cast_level(df$activity, nace_rev2, level = 1, label = FALSE))

Example of use within dplyr

df %>% 
  filter(cast_level(activity, level = 1, label = F) == "A") %>%
  group_by(activity = cast_level(activity, level = 3)) %>%
  summarise(some_total = sum(some_number))

complete_classification

To do

• Code is in package that passes R CMD check
• Write tests
• Utility functions, e.g. relevel, recode
• Different versions of labels, e.g. translations
• ...

Thank you, organisers!

Thank you, fellow participants!

... and thank you public for all of the suggestions that you are going to do

at https://github.com/uRos2018/categorical ;)

uRos2018/categorical documentation built on May 29, 2019, 9:15 a.m.