Skip to content

Creating Dummy Var in R for regression?

7 messages · Shivi Bhatia, Rui Barradas, Bert Gunter +2 more

Original
Shivi Bhatia
# 
Dear Team,

I need help with the below code in R:

gender_rec<- c('Dr','Father','Mr'=1, 'Miss','MS','Mrs'=2, 3)

reasons$salutation<- gender_rec[reasons$salutation].

This code gives me the correct output but it overwrites the
reason$salutation variable. I need to create a new variable gender to
capture gender details and leave salutation as it is.

i tried the below syntax but it is converting all to 1.

reasons$gender<- ifelse(reasons$salutation== "Mr" & reasons$salutation==
"Father","Male", ifelse(reasons$salutation=="Mrs" & reasons$salutation==
"Miss","Female",1))

Please suggest.
Rui Barradas
# 
Hello,

Your ifelse will never work because
reasons$salutation== "Mr" & reasons$salutation=="Father" is always FALSE
and so is reasons$salutation=="Mrs" & reasons$salutation=="Miss".
Try instead | (or), not & (and).

Hope this helps,

Rui Barradas

?

Citando Shivi Bhatia <shivipmp82 at gmail.com>:
?
Bert Gunter
# 
Just commenting on the email subject, not the content (which you have
already been helped with): there is no need to *ever* create a dummy
variable for regression in R if what you mean by this is what is
conventionally meant. R will create the model matrix with appropriate
"dummy variables" for factors as needed. See ?contrasts and ?C for
relevant details and/or consult an appropriate R tutorial.

Of course, if this is not what you meant, than ignore.

Cheers,
Bert


Bert Gunter

"The trouble with having an open mind is that people keep coming along
and sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )
On Fri, Aug 5, 2016 at 1:49 PM, <ruipbarradas at sapo.pt> wrote:
Shivi Bhatia
# 
Thanks you all for the assistance. This really helps.

Hi Bert: While searching nabble i got to know R with factors variables
there is no need to create dummy variable. However please consider this
situation:
I am in the process of building a logistic regression model on NPS data.
The outcome variable is CE i.e. customer experience which has 3 rating so
ordinal logistic regression will be used. However most of my variables are
categorical. For instance one of the variable is agent knowledge which is a
10 point scale.

This agent knowledge is again a 3 rated scale: high medium low hence i need
to group these 10 values into 3 groups & then as you suggested i can
directly enter them in the model without creating n-1 categories.

I have worked on SAS extensively hence found this a bit confusing.

Thanks for the help.
On Sat, Aug 6, 2016 at 2:30 AM, Bert Gunter <bgunter.4567 at gmail.com> wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      jeremiah rounds
    

    

      
      #
    

  


  

      
Something like:

d  =  data.frame(score = sample(1:10, 100, replace=TRUE))
d$score_t = "low"
d$score_t[d$score > 3] = "medium"
d$score_t[d$score >7 ] = "high"
d$score_t = factor(d$score_t, levels = c("low", "medium", "high"),
ordered=TRUE)  #set ordered = FALSE for dummy variables
X = model.matrix(~score_t, data=d)
X

        
On Fri, Aug 5, 2016 at 3:23 PM, Shivi Bhatia <shivipmp82 at gmail.com> wrote:

        

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        

  
  


  


      

    

    
1 day later

    

      
      

        


  

        

      Shivi Bhatia
    

    

      
      #
    

  


  

      
Thank you Jeremiah and all others for the assistance. This really helped.

On Sat, Aug 6, 2016 at 5:01 AM, jeremiah rounds <roundsjeremiah at gmail.com>
wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      Fredrik Karlsson
    

    

      
      #
    

  


  

      
Hi,

please also have a look at the 'cut' function.Very handa function for these
types of situations.

Best,

Fredrik

        
On Sun, Aug 7, 2016 at 8:10 PM, Shivi Bhatia <shivipmp82 at gmail.com> wrote: