On Thu, 19 Jun 2003, Dianne Worth wrote (referring to a Likert variable
whose values are -7, -6, ..., -1):
> Thank you -- because this is an X (as opposed to Y), I assume that
> it's appropriate to transform by x-squared.
It would be OK; I don't know about "appropriate". Notice that you'd
end up with the same set of values whether you were using a "positive"
item (+1,...,+7) or a "negative" item (-7,...-1).
> However, when I do that, the results simply appear to have become
> positively skewed. Am I missing something? DW
Nope. Only, the degree of skew will be somewhat different for the
squared values (1,4,9,...,49) than for the absolute values (1,...,7).
If I were doing it, I'd be inclined to use what you have called the
absolute values, equivalent to what I would call multiplying by -1 (that
is, reflecting the values about zero); because those values are so much
closer to the original ("raw") data. Unless, of course, there were some
inherent reason, or interpretation, related to the squared values.
(Your posts to date have not hinted at any such reason/interpretation.)
-- DFB.
-----------------------------------------------------------------------
Donald F. Burrill [EMAIL PROTECTED]
56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
