Re: st: 1-4 scale

[David Hoaglin <dchoaglin@gmail.com>]

Dear Ebru,

Your options probably depend on how much data you have.

In some areas, it is customary to assign the values 1 through 4 to
each question, form the sum of the four questions in each of the two
headings, and use those two variables in your model.  Centering is
probably not important at this stage.

That approach chooses one particular way of combining the four
questions in each heading.  To let the data choose the combination,
you can use all eight questions in your model, instead of the two
sums.  You should probably check that no two questions are highly
correlated.

If you have enough data, you can (as Nick suggested) use a factor
variable for each question.  Then your model will have three
coefficients for each question.  This approach avoids the assumption
that the response categories for a question are equally spaced.

I would not exclude the possibility of interactions between questions.
 In principle, you can cross-classify the data on the eight questions,
but the full cross-classification would have 4^8 = 2^16 cells.  For
the separate sets of four questions, the cross-classification would
have 256 cells.

David Hoaglin

On Sat, Aug 4, 2012 at 6:24 AM, Ebru Ozturk <ebru_0512@hotmail.com> wrote:
>
> Dear All,
>
> I have a question on how to add a variable that includes items measured by likert scale to the estimation model.
>
> For instance, in the data there is one question that measures the objectives of innovation activities. The answers to the question include 8 items that have been measured by 1-4 scale. I will split them into two headings as they focus on two types of objectives. Thus, for each 4 item how do I add them to the model? Because it is 1-4 scale so do I need to mean centred them or do something else?

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/