Date:         Mon, 23 Mar 2009 12:45:26 +0000
Reply-To:     Ruben van den Berg <ruben_van_den_berg@hotmail.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Ruben van den Berg <ruben_van_den_berg@hotmail.com>
Subject:      Confidence interval for 2 (related) percentages based on a single
              11 point Likert scale
Content-Type: multipart/alternative;


Dear all,

 

I've the following situation:

 

A 11-point Likert scale (from 0 (very dissatisfied) through 10 (very satisfied)) has been used in order to assess overall satisfaction. Now, the answers are recoded as follows:

 

RECODE NPS (0 thru 6=1) (7 thru 8=2) (9 thru 10=3) into SAT_GR.

EXE.

 

FRE SAT_GR.

 

Now the percentage (SAT_GR=1) is subtracted from the percentage (SAT_GR=3) (category 2 is not used here). The result of this is INDEX (with a theoretical range from -100% to +100%).

 

Can a confidence interval be computed for INDEX and if so, how?

 

I know there's a paired samples (difference between 2 proportions) test but these proportions are based on two different items (e.g. "do you have a cat (y/n)" versus "do you have a dog (y/n)"). I guess these formulas won't hold for proportions based on a single item...

 

Any help is highly appreciated!

 

TIA and kind regards,

 

Ruben v.d. Berg

_________________________________________________________________
What can you do with the new Windows Live? Find out
http://www.microsoft.com/windows/windowslive/default.aspx

[text/html]