What does "model" mean?

[From Bruce Nevin (11.11.2002 12:12 EST)]

The term "a model, the model" is used in CSG writings in two very different ways. This results in confusion.

1. "Model" means a specific implementation which, when run, displays behavior whose metrics conform to those measured for the organism(s) being modelled.

2. "Model" is used to mean something broader, perhaps control theory itself, of which models in sense (1) are instantiations or versions.

Because Model(2) is a class whose members are Models(1), equivocation between these two senses results in errors of logical typing. The least damaging effect of this is to confuse communication. Errors of logical typing can have much more serious effects as well.

The following recent exchange provides an example of this confusion:

> [From Rick Marken (2002.11.10.1030)]
>
> > Bill Williams (UMKC 10 November 2002 00:30 AM CST) --

[...]

> > A variety of "Giffen" type models have emerged.
>
> I think we have seen only one type of model: a control model.

OK. As you say below there are several different versions.

> I have seen two
> versions of the control model: one which controls for calories,
> expenditures and "prestige" and the other which controls for just
> calories and expenditures.

I feel very strongly that we should pick one and stick to it.

I personally prefer "Model" to mean something very specific: a particular model of specified behavioral data. Rather than talking about variants of a model, I think we should talk about related models.

If we continue to talk of a class of models(1) as a model(2) then we need to specify how comprehensive the class is. The above exchange refers to a very small class of two models in process of being specified. "The PCT model" of behavior as presented in B:CP, etc. is a class that includes an infinite number of models that have not yet been thought of, much less designed and specified.

         /Bruce Nevin

···

At 05:35 PM 11/10/2002, William Williams wrote:

[From Bill Powers (2002.11.11.1155 MST)]

Bruce Nevin (11.11.2002 12:12 EST)]

The term "a model, the model" is used in CSG writings in two very different
ways. This results in confusion.

I agree. The problem is that "model" has many popular meanings, of which
only one is the one we usually intend. I suggest using the term
"simulation" for a working model that actually produces predictions about
behavior that could be compared with real data.

Best,

Bill P.

[From Bill Williams UMKC 12 November 2002 4:00AM CST]

[From Bruce Nevin (11.11.2002 12:12 EST)]

The term "a model, the model" is used in CSG writings in two very different
ways. This results in confusion.

Agreed!

If we continue to talk of a class of models(1) as a model(2) then we need
to specify how comprehensive the class is. The above exchange refers to a
very small class of two models in process of being specified. "The PCT
model" of behavior as presented in B:CP, etc. is a class that includes an
infinite number of models that have not yet been thought of, much less
designed and specified.

Also Agreed.

Un-neccesary confusion might be avoided by using specific captions (GiffenV6)
for an implementation of a control theory model.

Best

Bill Williams

···

______________________________________________________________________
Do you want a free e-mail for life ? Get it at http://www.email.ro/