The Math Forum

Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Math Forum » Discussions » Courses » ap-stat

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: Likert
Replies: 0  

Advanced Search

Back to Topic List Back to Topic List  
Richard Scheaffer

Posts: 440
Registered: 12/6/04
Posted: Feb 28, 1997 12:37 PM
  Click to see the message monospaced in plain text Plain Text   Click to reply to this topic Reply

There seems to be a lot of interest in Likert scales
of late. Such scales (order of preference among 1,
2,..,5, etc.) produce data that should be considered
as ordered categorical. If comparing two or more
data sets of this type across nominal categories
(like gender), then rank procedures should be used.

If comparing across levels of another ordered
variable (such as grade level), procedures for
measuring association between ordered categorical
variables should be used. The best ones involve
concordant and discordant pairs, which are not
difficult to count but which are not mentioned in
introductory courses. (Perhaps they should be.)

If used as the response variable in a regression
setting, then logistic regression can be employed.
In the regression setting, treating the ordinal
categories as continuous measurements is not too bad
as long as there are at least 5 scale levels and a
reasonable amount of data.

In short, it is not a good idea to use Likert scale
data for illustrating the basic procedures covered in
an introductory course. Even the chi-square test
yields very little information here (as has been
pointed out) since it tests only for overall
association and not directions or strengths of
possibly interesting specific associations.

Dick Scheaffer

Point your RSS reader here for a feed of the latest messages in this topic.

[Privacy Policy] [Terms of Use]

© The Math Forum at NCTM 1994-2018. All Rights Reserved.