- My attitude regarding my article has not changed

one iota, despite appearances to the contrary based

on remarks by me that have been both taken out of

context and misunderstood.

Eric Laupot

____________________________________________________________________

Get free email and a permanent address at http://www.netaddress.com/?N=1 - In reply to Alfred M. Kriman's various recent

messages, let me say that he has presented us

with at least two serious contradictions:

1. He claims to have knowledge of statistics,

and he indeed often knows how to "walk the walk" and

"talk the talk." Yet his application of basic

statistical principles to the specific and rather

elementary problem at hand is so flawed as to present us with

a serious contradiction between his knowledge and

his actions. Specifically, Kriman's argument regarding

the instant statistical problem fails in at least two

respects:

A. Although Severus (or the author of fragment 2)

only chose *one* metaphor for the Christiani, Kriman

claims we should treat the problem mathematically

as if Severus had made 20 choices. This is nonsense,

as I have stated before.

B. Kriman claims that a Zipf distribution function

is applicable in the present case, which is likewise

not true. This is in part because we have no data

on how Severus or anyone else would have gone about

choosing their metaphor for "Christiani." The Zipf

distribution function purports to supply this data

by assuming that this choice would have been made

according to the words which were generally most popular

in Latin, but this is arguing from the general to

the specific in this case -- that is to say, we

simply have no data as to what metaphor

someone would have used for the word *Christiani* in

fragment 2. The Zipf distribution function doesn't

cover this contingency. *In addition,* upon a moment's

reflection one can see that the probability calculation

here involves a straight-up, one-time computation

involving a simple numerator and denominator, and

the Zipf function has nothing to do with this sort of

thing and wouldn't affect the calculation anyway.

Indeed, no distribution curve is applicable here,

including the Bell curve, because there is nothing to

distribute. Kriman also brings the Perseus

project into this, apparently implying they

have made some new discovery involving the Zipf

function that is applicable to fragment 2. This is

entirely misleading, as the Perseus project is not

working on the specific question of what metaphor

Severus or someone else would have used to describe

"Christiani." They are working on entirely different,

and more general, matters. This is like comparing

apples with oranges.

This disposes of Kriman's statistical claims.

2. Kriman has also presented us with a contradiction

in that he has strongly (and falsely) implied that in

effect I falsified research data by lying in my

article by claiming I had consulted the statistician,

Dr. Robert Gorman, when supposedly, according to

Kriman, I had not. The contradiction Kriman presents

us with involves the fact that Kriman never checked his

statement out with Dr. Gorman, despite the fact

that, virtually by Kriman's own admission, it would

have been very easy for him to have done so. For

by his own admission in his communication to

Classics-L of 11/13/00 (reposted to this group, for

reasons that are not entirely clear, by Jeffrey B.

Gibbons on 11/28/00), Kriman revealed, on one of his

Internet pop-ups, that Gorman is employed by Dade

Behring. The pop-up even gave Gorman's job title. It

would have been child's play to locate Gorman, but

Kriman did not do so. Also, Gorman's home town of

Elkton, Maryland was mentioned in the footnote in my

article that Kriman quoted from verbatim in his

message, and Gorman's home phone number is listed in

Directory Assistance and on the Internet. Yet Kriman

claimed he was unable to locate Gorman. There is a

huge inconsistency here.

Finally, I stand by my the argument in my

article and encourage everyone to seek an

independent opinion from a professional statistician

if they would like to know more about the

statistical issues involved.

Sincerely,

Eric Laupot

PO Box 286510

New York, NY 10128

USA

elaupot@...

Tel. (212) 744-9450

____________________________________________________________________

Get free email and a permanent address at http://www.netaddress.com/?N=1 - At 10:50 PM 12/1/00 -0500, Eric Laupot wrote:
>In reply to Alfred M. Kriman's various recent

I disagree. I think Kriman has a good point about the appropriateness of

>messages, let me say that he has presented us

>with at least two serious contradictions:

>

>... Kriman claims that a Zipf distribution function

>is applicable in the present case, which is likewise

>not true. This is in part because we have no data

>on how Severus or anyone else would have gone about

>choosing their metaphor for "Christiani." The Zipf

>distribution function purports to supply this data

>by assuming that this choice would have been made

>according to the words which were generally most popular

>in Latin, but this is arguing from the general to

>the specific in this case -- that is to say, we

>simply have no data as to what metaphor

>someone would have used for the word *Christiani* in

>fragment 2. The Zipf distribution function doesn't

>cover this contingency.

the Zipf distribution.

It is prudent to assume that the popular distribution prevails unless you

can show otherwise. You have not shown that some other choice model is more

likely; you merely speculate about different possibilities. Furthermore,

you admit that we lack data to determine the choice model. It would be

foolish to suppose that the choice of nouns is equiprobable, that is, that

each noun is equally likely to have been chosen-- I don't know of any

language that works like that. Under these conditions, the Zipf

distribution seems quite appropriate.

> *In addition,* upon a moment's

Maybe if one only devotes a moment's reflection; but if one thinks about it

>reflection one can see that the probability calculation

>here involves a straight-up, one-time computation

>involving a simple numerator and denominator, and

>the Zipf function has nothing to do with this sort of

>thing and wouldn't affect the calculation anyway.

a little longer, your argument does not hold water.

>Indeed, no distribution curve is applicable here,

This is nonsense, and shows a lack of mathematical intuition.

>including the Bell curve, because there is nothing to

>distribute.

>...This disposes of Kriman's statistical claims.

Au contraire.

Bob

Robert M. Schacht, Ph.D.

Northern Arizona University

Flagstaff, AZ

[Non-text portions of this message have been removed] - Bob Schacht believes he is on to something.

I wonder, therefore, if Bob wouldn't mind actually

running through his actual calculation for us, as I

did in my article. If he takes the time to do this,

I believe he will discover, as I stated, that it

cannot be done and that the Zipf function is

irrelevant here.

Regards,

Eric Laupot

____________________________________________________________________

Get free email and a permanent address at http://www.netaddress.com/?N=1