==================================================================
The gateway between this list and the sci.stat.edu newsgroup will
be disabled on June 9. This list will be discontinued on June 21.
Subscribe to the new list EDSTAT-L at Penn State using the web
interface at http://lists.psu.edu/archives/edstat-l.html.
==================================================================
.
On Fri, 04 Jun 2004 10:30:24 +0100, Thom <[EMAIL PROTECTED]>
wrote:
> Anders D Hojen wrote:
[snip]
... I take the listing of
> > alternatives to mean that the z-score approach is not preferable; maybe
> > particularly because it neutralizes within-group main effects (It was so
> > nice to get rid of those main effects, though). If anybody thinks that
> > there is NO problem with neutralizing certain main effects by using
> > z-scores, let them speak up now or forever hold there peace!
Getting rid of the nuisance-main effect is nice. That is one of the
hopes of doing the re-parameterization by standardizing by syllable.
Did that work?
The problem with the z-scores is that it is ad-hoc, and it could
be defined several ways (how many z-scores?), and it complicates
the explanation.
Using a log-transform was a basic idea for dealing with durations,
since the error is reasonably described as the percent change.
What is weak about it, here, is that it does *not* address the
nuisance main effect.
>
> I don't think it is a particular problem that the z transformation
> removes uninteresting main effects (ANOVA seems robust to similar
> problems for ipsative data such as ranks where the some means is
> constrained to add to a constant).
>
> I think the specific problem is that no-one else is convinced that the z
> transform directly addresses the syllable length problem.
I wonder if there are 'pauses' that might interfere with solving
the syllable length problem.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
