Hi,
first of all, thanks to all the people who gave me a hand with my previous
question on Chi-square (it really helped me to understand better what I
was doing).
But now I got seriously stuck again. I have looked at all the statistics
books and on-line help that I could get my hands on, but I'm not sure how
to proceed. If you can give me a hand with this one, I would really
appreciate it (I hope some day I can be on the helping side! :-))
I have a questionnaire with many questions, and each question with five
possible answers (Very low, low, average, high, very high).
1st problem
-----------
How could I measure whether people's replies tend to cluster
in some part of the scale?
Basically I want to find a way of eliminating the questions where
people's replies don't cluster around a section of the scale.
For instance, if for quetions a) there are 50 "very low" and 50 "very
high" and the rest 0, I would like to eliminate question a), as the
replies form two clear clusters.
If for question b) we get 40 "very low" and 60 "low" and the rest 0, I
would like to keep this one, as people's replies tend to cluster around
the lowest two values.
2nd problem
-----------
Let's say I found a nice solution for problem 1, so I have a set of
questions where most people's replies are grouped around 1 or 2 values in
the scale.
Now, for each of these questions I have a reply given by a computer
program. Let's say, for question b) my program says "very low".
How can I calculate my computer program's 'level of correctness' for this
question, compared with the answers gathered in the questionnaire?
===========
Sorry for the long e-mail, but I tried to make it as brief as possible,
while trying at the same time to make my problems clear.
If someone can give me a hand with these, I would appreciate it loads!
Thanks a lot,
Angel de Vicente
---------------------------------------------------------------
Learning Technology Developer
SELLIC
Edinburgh University
PhD Research Student
Artificial Intelligence
Division of Informatics
Edinburgh University
---------------------------------------------------------------
|