A Simple and Flexible Recoder

Description

Simple, easy to use, and quite flexible function for recoding variables. Among the features is that it can handle compound conditions and can include functional transformations on the go, which together allow for rapid piecewise definition of functions.

Details

Author(s)

Ali Sanaei

Simple and Handy Recoder

Description

A flexible function for recoding a variable. It can be used for anything from simple explicit recoding to piecewise-defined transformations.

Usage

recoder(var, recode, other = NULL, na = NA, as.what = NULL)

Arguments

Details

recode contains the instructions for how the recoding is done. Each instruction is in the form of condition: newvalue and instrunctions must be separated by semicolons (;). E.g., 1:2; 2:-1; 3:0 will change all ones to two, twos to negative one, and threes to zero. Instruction can include compound conditions and functions for newvalue expressed in terms of $. For example, "<0:0; >2:4; >=0 & <=2 : $^2" will give zero for all the values below 0, 4 for all the values above 2, and squared values for anything between 0 and 2.

The input can be other types beside numeric. For characters and factors, it is important that the form "old":"new" includes quotations around both the left-hand side and the right-hand side.

When the input is a factor, recoding works as expected, but new levels cannot be assigned. In order to assign new levels, the input vector should be made into something else (e.g., a character) and then the argument as.what could be set to "factor" to get a factor as the end result.

Value

The recoded vector is returned.

Author(s)

Ali Sanaei

See Also

cut, recode

Examples

u = rnorm(1000)
u.2 = recoder(u,'<-2: -2; >2 : 2' ,other = 0)
plot(u,u.2)

# use a function of 'x' for assigning new values based on old values
u.3 = recoder(u,'<-2: -2; >2 : 2; >=-2 & <=2 : $^3/4') 
# NOTE: instructions in recode are executed from left to right
# So u.3 and u.4 are the same
u.4 = recoder(u,'<-2: -2; >=-2 : $^3/4;  >2 : 2')
plot(u, u.3)

# another example of a piecewise defined function
x = runif(1000,-1,10)
y = recoder(x,'>0 & <1: sqrt($); >=1 : 1+log($)' ,other=NA) 
plot(y~x)
table(is.na(y))

# we can also create other types of variables
k = rpois(n = 10000, lambda = 3)
evodd = recoder(k %% 2, '0: "even"; 1: "odd" ')
table(evodd)

# the output could also be made a factor, etc., using "as.what" argument
evodd.factor = recoder(k %% 2, '0: "even"; 1: "odd" ', as.what='factor')

# characters are recoded similarly
str = c('a','b','x','z',NA) 
str2 = recoder(str, '"x":"c"; "z":"d" ')
str3 = recoder(str, '"a":1; "b":2; "x":3; "z":4 ')
print(str3) #note that this is still not numeric
num3 = recoder(str, '"a":1; "b":2; "x":3; "z":4 ', as.what='numeric')
print(num3)

# when input is a factor, the levels are fixed
f1 = factor( c('a','b','x','z',NA)  )
f2 = recoder( f1, '"x":"a"; "z":"b" ')

## The following line causes an issue
## because "c" and "d" are not recognized 
## as existing labels.
#  f3 = recoder(f1, '"x":"c"; "z":"d" ') 
# The proper way to do it is to first turn f1 
# into a character vector and then recode it.
f3 = recoder( as.character(f1), '"x":"c"; "z":"d" ' , as.what= "factor")