Re: determining if the difference in distributions of scores is statistically significant
- From: "dvschwab@xxxxxxxx" <dvschwab@xxxxxxxx>
- Date: 15 Jun 2005 11:12:15 -0700
Rupert,
You would use one of the many ANOVA (analysis of variance)
methods. I'm not sure from your question which of the differences you
are most interested in (that is, the difference between getting 5 and 4
right, 5 and 1 right, 4 and 0 right, etc; they are all different
questions). Also, if your unit of analysis is the group, then you will
need to administer the test to more than 4 groups to get a
statistically significant result; 25 would be the minimum I would feel
comfortable with. There are many good books on education statistics
that can tell you how to calculate an ANOVA for the questions you are
most interested in. Good luck!
David
Rupert wrote:
> Suppose you have a survey with 5 questions which you give to 4 groups
> and for each group you measure the distribution of who got 5 questions
> right, 4 questions right, 3 questions right, and so on. How would you
> determine whether the difference in distributions of scores is
> statistically significant?
.
- References:
- Prev by Date: Re: Cantor and the binary tree
- Next by Date: Re: Cantor and the binary tree
- Previous by thread: determining if the difference in distributions of scores is statistically significant
- Next by thread: Re: determining if the difference in distributions of scores is statistically significant
- Index(es):
Relevant Pages
|
