Writing as an Reply to my post, 
on 3 Dec 2003 10:52:07 -0800, [EMAIL PROTECTED] (SR Millis) wrote:

> As Senn, Stevens, & Chaturverdi (Statist Med, 2000,
> 19, 861-877) discuss, MANOVA doesn't provide useful
> estimates of the treatment effect. 

As Jim recommended, and I seconded, there
was a linear trend component:  That gives an
estimate of effect.


> 
> Better alternative include the summary measures
> approach (eg, regression through the origin), random

Huh?  What I think of as "regression through the origin"
is hardly ever useful for anything, especially, statistical
testing.  What was this referring to?

On the other hand, A linear trend is a compact
"summary measure"; and so is the Endpoint; and 
the Regressed change score (from covariance) -- I
thought that was where I was.

> coefficients, or generalized estimating equations.
> 
> SR Millis
[ snip, previous posts]

I  imagine those as giving *tests*  more readily than they 
give  estimates of effects.  But I haven't been using them.


-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
"Taxes are the price we pay for civilization." 
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to