Date: Dec 2, 2012 2:53 PM
Author: Halitsky
Subject: Re: Glad you brought up “singleton” length inter<br> vals ... been thinkin’ on ‘em also ...
I am shortly sending two csv files offline: a1_N_1_C.csv and

a1_N_1_S.csv, both with the column headers:

Fold

Frame

DiSet

SetTyp

LenID (this is now a ?singleton? length)

obsN

Mean u (actually, mean(u/(1+u))

Mean e

Auq (see below for definition of this column and the remaining

columns

AuqSE

Aubu

AubuSE

Aube

AubeSE

AubAvcov

RuqResVar

RubResVar

With respect to these columns, please note that:

Auq is the average slope of the regression Ruq = c on (u?,u?^2), where

u? = u/(1+u)

AuqSe is the SE for Auq

Aubu is the average slope of the u? coefficient of the regression Rub

= c on (u?, e, u?*e) ... note that u? is now first and e is now

second ... but as you?ve pointed out, everything is symmetrical so

this switch from our original Rub = c on (e, u?, u?*e) is purely

notational, i.e. either way, we?re interested in the average slope and

SE that I?ve called Aube, AubeSE in the attached files (see below),

and perhaps also the average slope and SE that I?ve called Aubu,

AubuSE (see below).

AubuSe is the SE for Aubu

Aube is the average slope of the e coefficient of the regression Rub =

c on (e, u?, u?*e)

AubuSe is the SE for Aube

AubAvcov is that potentially useful covariance for Rub, the one you

defined as:

cov[Av1,Av2] = cov[a1,a2] + var[a3]*mean_x1*mean_x2 +

cov[a1,a3]*mean_x1 + cov[a2,a3]*mean_x2, where 1= u?, 2 = e, and 3 =

u?*e.

RuqResVar = variance of the residuals of Ruq

RubResVar = variance of the residuals of Rub.

If you want me to send the equivalent files for each pair of cells as

they're computed, please let me know.

Also, as I progress thru the run, I will be computing the master file

of N's you've asked for, although if I send you pairs of csv's for all

pairs of cells, you'll have these N's readily available.