Data Dictionary

data_dictionary.R

# Generate data dictionary

DESCRIPTIONS <- list(mpg = "Miles Per Gallon")

# data dict function
data_dict <- function(data, descriptions = NA, print_table = FALSE){
  # Adapted from https://github.com/pw2/Data-Dictionary-Function/blob/master/Data%20Dictionary%20Function.R
  if(!require(pacman)) install.packages("pacman")
  pacman::p_load('tidyverse', 'psych', 'ggpubr', 'gridExtra', 'knitr', 'data.table')
  
  options(scipen = 999)
  
  # Get variable info and NAs
  var_info <- data.table::data.table(variable = names(data),
                                     variable_type = sapply(data, class),
                                     cardinality = sapply(data, function(x)
                                                            x %>%
                                                              purrr::keep(~ !is.na(.x)) %>%
                                                              unique() %>% 
                                                              length()),
                                     n_missing = sapply(data, function(y) 
                                       y %>% purrr::keep(~ is.na(.x)) %>%
                                         length()),
                                     example = sapply(data, function(x) 
                                       stringr::str_trunc(as.character(purrr::detect(x, ~!is.na(.x))), width = 20))
                                     )

   # get descriptive stats
   desc_stats <- data.frame(variable = names(data), 
                            describe(data)[c(2:5, 13, 8:10)], 
                            row.names = NULL)
   
   # Create the data dictionary
   d_dict <- merge(var_info, desc_stats, by = "variable", sort=FALSE)
   d_dict <- d_dict %>% dplyr::mutate_at(vars("mean":"range"), .fun = round, 2)
   
   # NA's for summary stats of variables not of class numeric or integer
   d_dict <- d_dict %>%
     dplyr::mutate(mean = ifelse(variable_type == "numeric" | variable_type == "integer", mean, ""),
                   sd = ifelse(variable_type == "numeric" | variable_type == "integer", sd, ""),
                   median = ifelse(variable_type == "numeric" | variable_type == "integer", median, ""),
                   se = ifelse(variable_type == "numeric" | variable_type == "integer", se, ""),
                   min = ifelse(variable_type == "numeric" | variable_type == "integer", min, ""),
                   max = ifelse(variable_type == "numeric" | variable_type == "integer", max, ""),
                   range = ifelse(variable_type == "numeric" | variable_type == "integer", range, ""))
   
   setDT(d_dict)
   # Merge in descriptions
   if (!is.na(descriptions)) {
     desc <- data.table::data.table(variable = names(descriptions), description = descriptions)
     d_dict <- merge(d_dict, desc, all.x = TRUE, on = "variable")
     d_dict[is.na(description) == TRUE, description := ""][]
     missing_vars <- d_dict$variable %>% purrr::discard(~.x %in% names(descriptions))
     if(length(missing_vars) > 0) {
       warning(paste("\nMissing variable descriptions for: \n\n", paste0(missing_vars, sep="", collapse="\n"), sep=""))
     }
   }
   d_dict
}

data_dict(mtcars, DESCRIPTIONS)