st: long or wide in survey data

[Andrea Bennett <mac.stata@gmail.com>]

Hi again,

I have a data set in wide format containing survey data. I want to  
find out what affects as certain set of knowledge related variables.  
Therefore I tend to generate a knowledge index based on the correct  
answers. Now, each individual was asked 3 knowledge questions related  
to one topic while also each individual was asked about 5 topic. So, I  
have 3x5 knowledge related answers.

Now the question: Would you recommend keeping the data set in wide  
format and generate an overall knowledge index or else would you  
reshape the data set to long and only generate a knowledge index for  
each topic (so basically treating each individual & topic combination  
as a single observation). When going for long I think I will have to  
use a random effects model on the basis of each individual and/or use  
the -vce(cluster individual)- option.

Maybe some of you have some experience on this and could share it?

Many thanks for your consideration,

Andrea
*
*   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/