R help - Apr 2009 - better way of recoding factors in data frame?

Hi all,

I apologize in advance for the length of this post, but I wanted to make sure I
was clear.

I am trying to merge two dataframes that share a number of rows (but some are
unique to each data frame). Each row represents a subject in a study. The
problem is that sex is coded differently in the two, including the way missing
values are represented.

Here is an example of the merged dataframe:
> myFrame2
???SubjCode SubjSex? ? ? ? ? Sex
1? ? ? sub1? ? ???M? ? ? ???<NA>
2? ? ? sub2? ? ???F? ? ? ???<NA>
3? ? ? sub3? ? ???M? ? ? ???Male
4? ? ? sub4? ? ???M? ? ? ???<NA>
5? ? ? sub5? ? ???F? ? ? ???<NA>
6? ? ? sub6? ? ???F? ? ???Female
7? ? ? sub7? ? ? ? ? ? ? ???<NA>
8? ? ? sub8? ? ? ? ? ? ? ???<NA>
9? ? ? sub9? ? ? ???Not Recorded
10? ? sub10? ? ? ???Not Recorded

I then apply the following:
> myFrame2$SubjSex <- factor(myFrame2$SubjSex, levels =
c('M','F'))
> myFrame2$SubjSex <- factor(myFrame2$SubjSex, labels =
c('Male','Female'))
> myFrame2 <- transform(myFrame2, newSex = ifelse(is.na(SubjSex), Sex,
SubjSex))
...and get this:
> myFrame2
???SubjCode SubjSex? ? ? ? ? Sex newSex
1? ? ? sub1? ? Male? ? ? ???<NA>? ? ? 1
2? ? ? sub2? Female? ? ? ???<NA>? ? ? 2
3? ? ? sub3? ? Male? ? ? ???Male? ? ? 1
4? ? ? sub4? ? Male? ? ? ???<NA>? ? ? 1
5? ? ? sub5? Female? ? ? ???<NA>? ? ? 2
6? ? ? sub6? Female? ? ???Female? ? ? 2
7? ? ? sub7? ? <NA>? ? ? ???<NA>? ???NA
8? ? ? sub8? ? <NA>? ? ? ???<NA>? ???NA
9? ? ? sub9? ? <NA> Not Recorded? ? ? 3
10? ? sub10? ? <NA> Not Recorded? ? ? 3

I need that last column to have just 1 (Male), 2 (Female) or 0 (Missing), and
the only way I've come up with seems very kludgy:
> myFrame2$newSex[is.na(myFrame2$newSex)] <- 0
> myFrame2$newSex <- ifelse(myFrame2$newSex == 3, 0, myFrame2$newSex)
That gives me the right values for "newSex", but I'd like to
positively select for the values I want to keep, rather than negatively
selecting the ones to change - I tried this:
> myFrame2$newSex <- ifelse(myFrame2$newSex ==1 || myFrame2$newSex == 2,
myFrame2$newSex, 0)
But I just get 1 for every row in newSex. Does anyone know of a way to do this
by positively selecting the values 1 and 2?


Thanks,
Mohinder

On Thu, 9 Apr 2009 mohinder_datta at yahoo.com wrote:
>
> Hi all,
>
> I apologize in advance for the length of this post, but I wanted to make
sure
> I was clear.
Good strategy.

>I tried this:
>
>> myFrame2$newSex <- ifelse(myFrame2$newSex ==1 || myFrame2$newSex ==
2, myFrame2$newSex, 0)
>
First, you need |, not ||, because || returns just a single value.

This still won't quite work, because the condition will be NA when newSex is
NA (if you don't know a number then you don't know whether it is equal
to 1 or 2).

So,
myFrame2$newSex <- ifelse(myFrame2$newSex %in% c(1,2), myFrame2$newSex, 0)

or, slightly shorter
myFrame2$newSex <- with(myFrame2, ifelse(newSex %in% c(1,2), newSex, 0)

     -thomas

Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle

Thank you for your help!


--- On Thu, 4/9/09, Thomas Lumley <tlumley at u.washington.edu> wrote:
> From: Thomas Lumley <tlumley at u.washington.edu>
> Subject: Re: [R] better way of recoding factors in data frame?
> To: mohinder_datta at yahoo.com
> Cc: r-help at r-project.org
> Date: Thursday, April 9, 2009, 2:10 PM
> On Thu, 9 Apr 2009 mohinder_datta at yahoo.com
> wrote:
> 
> >
> > Hi all,
> >
> > I apologize in advance for the length of this post,
> but I wanted to make sure 
> > I was clear.
> 
> Good strategy.
> 
> 
> >I tried this:
> >
> >> myFrame2$newSex <- ifelse(myFrame2$newSex ==1
> || myFrame2$newSex == 2, myFrame2$newSex, 0)
> >
> 
> First, you need |, not ||, because || returns just a single
> value.
> 
> This still won't quite work, because the condition will be
> NA when newSex is NA (if you don't know a number then you
> don't know whether it is equal to 1 or 2).
> 
> So,
> myFrame2$newSex <- ifelse(myFrame2$newSex %in% c(1,2),
> myFrame2$newSex, 0)
> 
> or, slightly shorter
> myFrame2$newSex <- with(myFrame2, ifelse(newSex %in%
> c(1,2), newSex, 0)
> 
> ? ???-thomas
> 
> Thomas Lumley??? ???
> ??? Assoc. Professor, Biostatistics
> tlumley at u.washington.edu???
> University of Washington, Seattle
> 
> 
>