R help - Jun 2010 - Factoring a variable

Hi,

I have a dataset where the results are coded ("yes", "no") 
We want to
do some machine learning with SVM to predict the "yes" outcome

My problem is that if I just use the as.factor function to convert, then
it reverses the levels.

----------------------
x <- c("no", "no", "no", "yes",
"yes", "no", "no")
 as.factor(x)
[1] no  no  no  yes yes no  no
Levels: no yes
----------------------
The SVM function (in the e1071 package) sees "no" as the first label
and
treats that as the positive outcome. 
The problem arises when we look at the decision values of the
predictions.  Everything is gauged as values for "no". 

So, is there a way to force R to use my specified order when converting
to factors? 

I've tried as.factor(x, levels=c("yes", "no")) but that
throws errors
about unused arguments.

Any help?


Thanks

Hi,

I have a dataset where the results are coded ("yes", "no") 
We want to
do some machine learning with SVM to predict the "yes" outcome

My problem is that if I just use the as.factor function to convert, then
it reverses the levels.

----------------------
x <- c("no", "no", "no", "yes",
"yes", "no", "no")
 as.factor(x)
[1] no  no  no  yes yes no  no
Levels: no yes
----------------------
The SVM function (in the e1071 package) sees "no" as the first label
and
treats that as the positive outcome. The problem arises when we look at
the decision values of the
predictions.  Everything is gauged as values for "no".
So, is there a way to force R to use my specified order when converting
to factors?
I've tried as.factor(x, levels=c("yes", "no")) but that
throws errors
about unused arguments.

Any help?


Thanks

Noah Silverman wrote:
> Hi,
> 
> I have a dataset where the results are coded ("yes",
"no")  We want to
> do some machine learning with SVM to predict the "yes" outcome
> 
> My problem is that if I just use the as.factor function to convert, then
> it reverses the levels.
> 
> ----------------------
> x <- c("no", "no", "no", "yes",
"yes", "no", "no")
>  as.factor(x)
> [1] no  no  no  yes yes no  no
> Levels: no yes
You want to use ?factor, not ?as.factor.  Then you can use the
"levels"
argument.

factor(x, levels = c("yes", "no"))

Also see ?relevel.

Hi,

On Thu, Jun 17, 2010 at 3:45 PM, Noah Silverman <noahsilverman at
ucla.edu> wrote:
> Hi,
>
> I have a dataset where the results are coded ("yes",
"no") ?We want to
> do some machine learning with SVM to predict the "yes" outcome
>
> My problem is that if I just use the as.factor function to convert, then
> it reverses the levels.
>
> ----------------------
> x <- c("no", "no", "no", "yes",
"yes", "no", "no")
> ?as.factor(x)
> [1] no ?no ?no ?yes yes no ?no
> Levels: no yes
> ----------------------
> The SVM function (in the e1071 package) sees "no" as the first
label and
> treats that as the positive outcome.
> The problem arises when we look at the decision values of the
> predictions. ?Everything is gauged as values for "no".
>
> So, is there a way to force R to use my specified order when converting
> to factors?
>
> I've tried as.factor(x, levels=c("yes", "no")) but
that throws errors
> about unused arguments.
>
> Any help?
Yes, look at the error message you're getting in your call to
`as.factor` more closely -- also look at the help for as.factor and
note there is no "levels" argument:

R> x <- c("no", "no", "no", "yes",
"yes", "no", "no")
R> factor(x, levels=c('yes', 'no'))
[1] no  no  no  yes yes no  no
Levels: yes no

-steve

-- 
Steve Lianoglou
Graduate Student: Computational Systems Biology
 | Memorial Sloan-Kettering Cancer Center
 | Weill Medical College of Cornell University
Contact Info: http://cbio.mskcc.org/~lianos/contact