R help - Jan 2005 - Nested ifelse - is there a better way?

Dear r-help,

I'm interested in finding a better way to add a column to a data frame
when calculating the new column requires more than one conditional.

For example, if I wanted to associate a character string in
{"Pos","Neg","Zero"} with each number in the
following data frame:
> d <- data.frame(num = -2:2)
> d
  num
1  -2
2  -1
3   0
4   1
5   2

I could use a nested ifelse:
> d$p1 <- ifelse(d$num < 0, "Neg", ifelse(d$num == 0,
"Zero", "Pos"))
> d
  num   p1
1  -2  Neg
2  -1  Neg
3   0 Zero
4   1  Pos
5   2  Pos

which I believe becomes cumbersome if more conditionals are required.
I've also considered using sapply:
> d$p2 <- sapply(d$num, function(num) { if (num < 0) { "Neg"
} else if (num == 0) { "Zero" } else { "Pos" } })
> d
  num   p1   p2
1  -2  Neg  Neg
2  -1  Neg  Neg
3   0 Zero Zero
4   1  Pos  Pos
5   2  Pos  Pos

Is there a more elegant approach to situations like this?

Thanks in advance,

Jeff

Jeff> I'm interested in finding a better way to add a column to a data 
Jeff> frame when calculating the new column requires more than one
Jeff> conditional.

[description of doing this with lots of ifelse's deleted]

Jeff

If all you are doing is slicing one quantitative variable into categories, 
then see

 	?cut

and

 	?findInterval

---------

If you need to use multiple variable to define the splits, you might 
construct an index using a 'mixed radix positional numbering system' (I 
suggest you google that term if it isn't familiar)

As an example if you have two variables (x and y,say) coded as (0,1,2) and 
(0,1,2,3,4), then

 	z <- x * ( max(y) + 1 ) + y

gives you z with categories (0,...,14). Extended to three or more 
variables, you need to do some nested multiplications.

Alternatively, you can use tapply to get this type of result

 	z <- tapply( x, list( x, y ) )

but here the categories will be numbered from 1. Some care may be needed 
to be sure that the category numbers are what you wanted.

----------

Subscripting can be used to combine categories

Here I combine 6 and 7, 8 and 9, and 14 and 15:

 	z.recode <- c(1,2,3,4,5,6,6,7,7,8,9,10,11,12,12)

 	z2 <- z.recode[ z ]

and do not miss out on

 	?levels

as "levels(x) <- value" will do the same when x is a factor.

HTH,

Chuck

Charles C. Berry                        (858) 534-2098
                                          Dept of Family/Preventive Medicine
E mailto:cberry at tajo.ucsd.edu	         UC San Diego
http://hacuna.ucsd.edu/members/ccb.html  La Jolla, San Diego 92093-0717