Date: Dec 14, 2012 4:00 PM
Author: Halitsky
Subject: Response to your last of 12/14 at 227pm re terminology and methodology.

1.  You asked:

For instance, do you do it over Length for each cell in a 4-way (Fold
x Set x Subset x Method) table?

Exactly. Same as for the regression of Aubqe on L, as per my
explanatory post of 12/12 at 9:10 (which you readily understood.)

2. You wrote:

?there is another formula for the denominator of the t.?

Thanks as always for your time and patience in providing the details ?
I can readily compute the denominator you suggest because of what I
did to get the covariance matrix from Ivo?s module ? remember ? he
does the heavy lifting by returning an A that has been computed both
slickly and accurately. I do the rest, which is not as problematic.

3. You wrote:

?So far the all the plots look pretty linear, with just a hint of
positive curvature.?

Please see my post immediately following this one.

4. You asked:

"please explain the labels AuqSE, AubuSE, AubquSE, AubeSE, AubqeSE".

Sorry ? these labels come from what I?m calling the regressions
themselves (at least in my mind):

Ruq is the regression of c on (u,u^2), so Auq is the average slope as
defined for you for Ruq

Rub is the regression of c on (e,u,u*e), so AubuSE and AubeSE are the
two average slopes you defined for Rub ? one for u and one for e

Rubq is the regression of c on (e,u,u*e,u^2), so AubqeSE and AubquSe
are the two average slopes you defined for Rubq ? again, one for u and
one for e.

In all the above, ?q? signals the involvement of the square term ?u^2?
and ?b? signals the involvement of the cross-term u*e.

If you want to change notation here, please feel free to suggest
alternatives of your liking.

5.

You said:

please include the phrase "the regression of"

Will do.