[EpiData-list] Re: Ranked variable questions

epidata-list at lists.umanitoba.ca epidata-list at lists.umanitoba.ca
Thu Aug 23 20:33:09 CDT 2007


An important question is whether a response of 1 most important or least 
important?
If 1 is most important, then make the default value 11 and not zero.

I would just do:

describe v1-v25

and scan the medians to find those with the lowest median (if 1 is most 
important and default is 11) or highest median (10 is most important and 
default is 0). Items that are infrequently ranked will have a median 
equal to the default. Items most frequently ranked will have a median 
that is based mostly on ranks.

I haven't thought about the formal statistical analysis here, but you'll 
get a pretty good assessment.

Jens approach works very well when the most important response is 10. If 
most important is 1, then you can also do this:

v1 = 10-v1
v2 = 10- v2
etc
and use aggregate.

jamie
> Christine wrote:
> I have been asked to set up a database for a questionnaire that has 
> already been written. I was going to set it up in epidata but my 
> problem is how to deal witha question that asks people to rank 
> options. The question gives a list of topics and asks the person 
> filling in the  questionnaire to rank them from 1-10 in order of 
> importance.
>
Jens wrote:
> I suggest to do the following:
> For each question (say 25 different themes) enter the priority the 
> person gave to that issue.
> setup the qes:
> v1 item1 ..... ##
> v2 item2 ..... ##
> etc
> v25 text for item 25 ##
>
> in chk make the defaultvalue 0:
> before file
>   DEFAULTVALUE v1-v25 0
> end
>
> Then in analysis after entering the data: (say you had the 25 variables)
> read response
> aggregate /sum="v1,v2,v3,v4,v5"
> aggregate /sum="v6,v7,v8,v9,v10"   etc
>
> * the tables will show the sum of the variables, that is the overall 
> ranking sum.
>



More information about the EpiData-list mailing list