[R] See the numeric codes of a factor

Angel Rodriguez angel.rodriguez at matiainstituto.net
Fri Sep 19 14:55:42 CEST 2014

Well, a variable with values 0/1 is useful for calculating observed probabilities by groups. But it is not diffcult to have the same variable both as numeric and as a factor in the dataframe and use each variation depending on the analysis.


De: Duncan Murdoch [mailto:murdoch.duncan at gmail.com]
Enviado el: vie 19/09/2014 14:22
Para: Angel Rodriguez; r-help at r-project.org
Asunto: Re: [R] See the numeric codes of a factor

On 19/09/2014 8:12 AM, Angel Rodriguez wrote:
> Re: [R] See the numeric codes of a factor
> Thank you, Duncan. So isn't it possible to add labels to a variable
> with numeric values 0/1? This kind of variable is very useful for
> logistic regression, for example, but I'd rather have its
> categories labelled.

I think you are thinking of how you have done things in some other
system.  In R, a factor is fine in logistic regression, regardless of
the fact that internally values are stored as 1 and 2.

Duncan Murdoch

> Angel
> ------------------------------------------------------------------------
> *De:* Duncan Murdoch [mailto:murdoch.duncan at gmail.com]
> *Enviado el:* vie 19/09/2014 13:32
> *Para:* Angel Rodriguez; r-help at r-project.org
> *Asunto:* Re: [R] See the numeric codes of a factor
> On 19/09/2014, 6:53 AM, Angel Rodriguez wrote:
> > Dear Subscribers,
> >
> > I want to label a numeric variable 0="Bad" /1="Good". I understand
> the only way is to transform it into a factor variable.
> >
> > Is there a way to check that the numeric values of the new factor
> variable are 0 and 1 and not 1 and 2?
> If you apply as.numeric() to a factor, you won't get a zero value.
> Internal factor values start at 1.
> So I wouldn't rely on the internal storage to achieve whatever it is you
> want to achieve.  Use explicit computation, e.g.
> words <- ifelse(var == 0, "Bad", ifelse(var == 1, "Good", NA))
> values <- ifelse(words == "Bad", 0, ifelse(words == "Good", 1, NA))
> Duncan Murdoch

	[[alternative HTML version deleted]]

More information about the R-help mailing list