Date:         Wed, 27 Feb 2008 13:48:55 -0500
Reply-To:     "Long, Stuart (NIH/NIEHS) [C]" <long3@NIEHS.NIH.GOV>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Long, Stuart (NIH/NIEHS) [C]" <long3@NIEHS.NIH.GOV>
Subject:      Re: How should I code this problem?
Comments: To: Reeza <fkhurshed@HOTMAIL.COM>
In-Reply-To:  A<2bacbc46-b98c-4cb6-9eae-7cbd1a094540@e25g2000prg.googlegroups.com>
Content-Type: text/plain; charset="us-ascii"

The ordering of values in PROC LOGISTIC is done by the encoded values.
If the variable is not encoded, then the ordering is based on the actual
values of the variable.  I often remove the formatting if the actual
values have the orders that I want:

For a variable, say, truant(how many days missed)

VALUE 1 = "0-4 days"
      2 = "5-9 days"
      3 = "10-20 days"
      4 = "21+ days" ;

The CLASS order in the modeling will be

0-4 days
10-20 days
21+ days
5-9 days

Two solutions to this are:

1)Get rid of encoded values with
FORMAT truant;

OR

2)Make sure the encoded values are in the order you want:
VALUE 1 = "1) 0-4 days"
      2 = "2) 5-9 days"
      3 = "3) 10-20 days"
      4 = "4) 21+ days" ;


Stuart Long
Contractor
Westat
Senior Analyst
919-941-8307





-----Original Message-----
From: Reeza [mailto:fkhurshed@HOTMAIL.COM]
Sent: Wednesday, February 27, 2008 1:36 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Re: How should I code this problem?


On Feb 27, 11:21 am, Melodyp <pearsonmel...@gmail.com> wrote:
> Hello!
>
> I have a file with 589 varialbes, and in order to do the Logistic
> regression, I have to transfer many of the variables with letters in
> to numbers. Many of them contains the same results : such as N-->>No,
> Y-->> Yes, is there a way I can code them all together to let N-->>"0"
> and Y-->>"Y"?
>
> Thank you in advance for your help.
>
> Regards,
>
> Melody

Use a format and apply the format to all the variables you need. Though
I'm not 100% sure how the regression would work, would it use the
recoded or original value.