Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: Converting ratings to rankings


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: Converting ratings to rankings
Date   Tue, 20 Jun 2006 09:35:11 +0100

7 -> 1, 6 -> 2, etc. is in Stata terms 

8 - rank

if that is part of the question. A hand-waving argument 
will convince you that this is true. 

Subtracting 1/2 in the event of ties seems to follow 
in your case from (1) inverting the scale and (2) adjusting 
for cases in which two things are assigned the 
same rank. But it is not quite right for ties of 
three or more. 

The more general adjustment if you need 
to ensure that each respondent has the same total
weight in whatever you are doing is to assign to 
tied ranks the average of the ranks that would have been 
assigned otherwise. In Stata, this is simply done
by using -egen, rank()- on the "ranks". -egen, rank()-
has various options, but that is the default. 

So you need to flip the scale and massage the ties, 
and it doesn't matter which you do first. 

I don't see that you need to use -if- (or with 
a nod to my co-author David Kantor, -cond()-) 
for any of this. 

Suppose your variables are rank1-rank8. 

foreach v of var rank1-rank8 { 
	tempvar work 
	gen `work' = 8 - `v' 
	egen new`v' = rank(`work') 
	label var new`v' "corrected rank from `v'" 
	drop `work' 
} 

Nick 
n.j.cox@durham.ac.uk 

Mike Stern
 
> I have a programming question:
> 
> I have need to convert 8 variables from a social survey that 
> respondents
> rated on a scale from 1 to 7 to implied rankings.  The survey 
> question asks
> respondents to rate 8 items relative to one issue with 1 
> being much less
> important and 7 being much more important with 4 being just 
> as important
> (conceptual middle point).  Therefore, a ranking needs to be 
> relative to the
> center point and because it now needs to be a ranking 7 needs 
> to be recoded
> to 1, 6 needs to be recoded to 2 and so on.
> 
> This would not be a particularly thorny issue except for the fact that
> respondents often rated two or more of the variables at the 
> same level (e.g.
> two or three items given a 5).  For such ratings, I am taking 
> a 1/2 point
> off their ranking.  I have a situation where if respondents 
> rated two items
> 5 (after the recoding needed to convert a rating to ranking--much more
> important now equaling 1 instead of 7 and much less important 
> now equaling 7
> instead of 1 and so on) the two items receive a score of 3.5.
> 
> Does anyone know of ways that stata can convert ratings to 
> rankings without
> a large number of "if then" statements for every possible 
> combination of
> duplicate ratings?

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index