Date: Wed, 2 May 2007 08:17:14 -0400
Reply-To: Peter Flom <email@example.com>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Peter Flom <peterflomconsulting@MINDSPRING.COM>
Subject: Re: rescaling categorical survey data
Content-Type: text/plain; charset=UTF-8
Sekxanh Chanthaboury <schant508@HOTMAIL.COM> asked
>Does anyone know of the correct way to rescale categorical survey data? I
>have old data that is ranked from 1-4, but the new survey data (same
>questions) is scaled from 1-7. How do I correctly map the 1-4 values to
>the new 1-7 values?
There is, AFAIK, no single answer to this. And rescalings can make a difference. Back in grad school, I compared two scalings of the Beck Depression Inventory, and, while the clinical results were very similar, the factor structure changed.
OK. I am going to assume that the data is ordinal, but not interval. That seems reasonable, given how you have worded the question, although you should confirm that.
Then, I think you have to make the much more dangerous assumption that the people who took the survey with the 4 point scale are similar to those who took it with the 7 point scale, on whatever it is the survey was attempting to measuer.
Next, look at the proportion of people who gave each answer on each survey.
If you have other variables, and if N is large, you could try some sort of matching.
Interesting set of issues - could you write back to SAS-L with a little more detail?