Working with Code Lists.

codelist

codelist is a package that should make it easier to work with code lists. A code list is a set of codes with associated values. For example one could have codes ‘E’ = ‘Employed’, ‘U’ = ‘Unemployed’, ‘N’ = ‘Not belonging to the working population’ and ‘X’ = ‘Unknown’. Codes can also have a description, be contained in a hierarchy and indicate specific types of missing values.

Quick Overview

library(codelist)

First we define the code list. In this case it is defined manually; in practice these will often be read from file. A code list has at minimum codes and corresponding labels. Here we also have a column indicating which codes can be interpreted as missing values.

cl <- codelist(
  codes = c("E", "U", "N", "X"), 
  labels = c("Employed", "Unemployed", 
    "Not belonging to working population", "Unknown"),
  missing = c(0, 0, 0, 1)
)

We can use this code list to define a ‘coded’ vector; which is a vector of codes with an attribute ‘codelist’.

x <- coded(c("N", "E", "E", "U", NA, "E", "N", "X", "N"), cl)
x
## [1] N    E    E    U    <NA> E    N    X    N   
## 4 Codelist: E(=Employed) ...X(=Unknown)

The general idea of the codelist package, is that we work with the codes as these are generally the most accurate. However, for presentation and statistical analyses we will often want to work with the labels. The labels methods transforms the vector into a factor for analysis and presentation:

table(x)
## x
## E N U X 
## 3 3 1 1 

table(labels(x, missing = FALSE), useNA = "ifany")
## 
##                            Employed                          Unemployed 
##                                   3                                   1 
## Not belonging to working population                             Unknown 
##                                   3                                   1 
##                                <NA> 
##                                   1 

The code list can also be used to check if codes are valid making the code more safe:

try( x[1] <- "A" ) 
## Error in `[<-.coded`(`*tmp*`, 1, value = "A") : 
##   Invalid codes used in value.

try( any(x == "B") )
## Error in Ops.coded(x, "B") : Invalid codes used in RHS

In this case the codes are somewhat readable. However, generally when reading code it is difficult to understand what a line of code like the lines above means. For someone reading the code it is easier to work with the labels:

x[1] <- as.label("Employed")
x[is.missing(x)] <- as.label("Unemployed")

Of course using invalid labels will generated an error.

More information

More information can be found in the vignettes of the package:

• Introduction to working with code lists.