recode {memisc} | R Documentation |
recode
substitutes old values of a factor or a numeric
vector by new ones, just like the recoding facilities in some
commercial statistical packages.
## S4 method for signature 'vector': recode(x,...,otherwise="NA") ## S4 method for signature 'factor': recode(x,...,otherwise="NA") ## S4 method for signature 'item': recode(x,...,otherwise="NA")
x |
An object |
... |
One or more assignment expressions, each
of the form new.value <- old.values .
new.value should be a scalar numeric value
or character string. If one of the new.value s
is a character string, the return value
of recode will be a factor and each new.value
will be coerced to a character string that labels a level of the factor.
Each old.value in an assignment expression may be a
(numeric or character) vector. If x is numeric such an
assignment expression may have the form new.value <- range(lower,upper)
In that case, values between lower and upper are exchanged by
new.value . If one of the arguments to range is min ,
it is substituted by the minimum of x .
If one of the arguments to range is max ,
it is substituted by the maximum of x .
In case of the method for labelled vectors, the tags of
arguments of the form tag = new.value <- old.values
will define the labels of the new codes.
If the old.values of different assignment expressions overlap,
an error will be raised because the recoding is ambigous.
|
otherwise |
a character string or some other value
that the result may obtain. If equal to NA or "NA" ,
original codes not given an explicit new code are recoded into
NA . If equal to "copy" ,
original codes not given an explicit new code are copied.
|
recode
relies on the lazy evaluation mechanism of R:
Arguments are not evaluated until required by the function they are given to.
recode
does not cause arguments that appear in ...
to be evaluated.
Instead, recode
parses the ...
arguments. Therefore, although
expressions like 1 <- 1:4
would cause an error action, if evaluated
at any place elsewhere in R, they will not cause an error action,
if given to recode
as an argument. However, a call of the
form recode(x,1=1:4)
, would be a syntax error.
If John Fox' package "car" is installed, recode
will also be callable
with the syntax of the recode
function of that package.
A numerical vector, factor or an item
object.
recode
of package car
.
x <- as.item(sample(1:6,20,replace=TRUE), labels=c( a=1, b=2, c=3, d=4, e=5, f=6)) print(x) # A recoded version of x is returned # containing the values 1, 2, 3, which are # labelled as "A", "B", "C". recode(x, A = 1 <- range(min,2), B = 2 <- 3:4, C = 3 <- range(5,max), # this last comma is ignored ) # This causes an error action: the sets # of original values overlap. try(recode(x, A = 1 <- range(min,2), B = 2 <- 2:4, C = 3 <- range(5,max) )) recode(x, A = 1 <- range(min,2), B = 2 <- 3:4, C = 3 <- range(5,6), D = 4 <- 7 ) # This results in an all-missing vector: recode(x, D = 4 <- 7, E = 5 <- 8 ) f <- as.factor(x) x <- as.integer(x) recode(x, 1 <- range(min,2), 2 <- 3:4, 3 <- range(5,max) ) # This causes another error action: # the third argument is an invalid # expression for a recoding. try(recode(x, 1 <- range(min,2), 3:4, 3 <- range(5,max) )) # The new values are character strings, # therefore a factor is returned. recode(x, "a" <- range(min,2), "b" <- 3:4, "c" <- range(5,6) ) recode(x, 1 <- 1:3, 2 <- 4:6 ) recode(x, 4 <- 7, 5 <- 8, otherwise = "copy" ) recode(f, "A" <- c("a","b"), "B" <- c("c","d"), otherwise="copy" ) recode(f, "A" <- c("a","b"), "B" <- c("c","d"), otherwise="C" ) recode(f, "A" <- c("a","b"), "B" <- c("c","d") )