Ted: You are either being deliberately obtuse or playing Devil's
advocate or just stirring. It is clear from his/her posts that the OP
has limited understanding of both R and statistics. Your sophisticated
philosophising about the possibility of "three sexes" is very unlikely
to have anything to do with what the OP wishes to accomplish.
The advice requested was not about how to treat "NA" as a "third sex"
but about how to convert categorical data coded as NA, "M", and "F" to
numeric. Which cannot possibly be a good idea.
It is not productive to encourage the OP to persist with wrong-headed
notions. Instead he or she should be encouraged to get to grips with
the real issues of the analysis and to understand that treating
categorical data as numeric is a recipe for disaster.
cheers,
Rolf
--
Technical Editor ANZJS
Department of Statistics
University of Auckland
Phone: +64-9-373-7599 ext. 88276
On 01/11/15 08:47, Ted Harding wrote:
[Apologies if the message below should arrive twice. When first
sent there was apparently something wrong with the email address
to r-help, and it was held for moderation because "Message has
implicit destination" (whatever that means). I have made sure
that this time the email address is correct.]
John Fox has given a neat expression to achieve the desired result!
I would like to comment, however, on the somewhat insistent criticism
of Val's request from several people.
It can make sense to have three "sex"es. Suppose, for example,
that the data are records of street crime reported by victims.
The victim may be able to identify the sex of the preprator
as definitely "M", or definitely "F". One of the aims of the
analysis is to investgate whether there is an association
between the gender of the offender and the type of crime.
But in some cases the victim may not have been able to recognise
the offender's sex. Then it would have to go in the record as "NA"
(or equivalent). There can be two kinds of reason why the victim
was unable to recognise the sex. One kind is where the victim
simply did not see the offender (e.g. their purse was stolen
while they were concentrating on something else, and they only
found out later). Another kind is where the offender deliberately
disguises their gender, so that it cannot be determined from their
appearance. This second kind could be associated with a particular
category of crime (and I leave it to people's lurid imaginations
to think of possible examples ... ).
Then one indeed has three "sex"es: Male, Female, and Indeterminate,
for each of which there is a potential assoctiation with type of crime.
With most analyses, however, a category of "NA" would be ignored
(at least by R).
And then one has a variable which is a factor with 3 levels, all
of which can (as above) be meaningful), and "NA" would not be
ignored.
Hoping this helps to clarify! (And, Val, does the above somehow
correspond to your objectives).
Best wishes to all,
Ted
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
