On 08/23/2011 08:39 PM, Iasonas Lamprianou wrote:>
>
> Dear colleagues,
>
>
> I would like to run a cluster analysis on a number of variables. They are
Likert Scale (0 to 10), but they also have a "Don't know' option at
the end of the scale. Apparently, with the 'Don't Know' option in
place, they cannot be considered to be linear or ordinal. How can these
variables be analysed using R? One option would be to use poLC and treat them as
categorical variables to classify the respondents into clusters (latent
categorical variable). But is there a way to treat them as partly
ordinal/conitinuous and partly categorical?
>
> Thank you
> Jason
>
Hi Jason,
The way I have treated this sort of response measure is to create two
variables, as two questions are being asked. One is:
"Do you have an opinion on this statement?"
and the other is:
"How much do you agree or disagree with this statement?"
(if the item really is a Likert scale). A "No" on the first question
automatically implies a missing value on the second.
Oh, and I usually tell the person who has written the questionnaire to
limit each item to one question.
Jim