Logit/Probit Models

Logit/Probit Models

Post by Matthe » Wed, 23 Jul 2003 04:01:30



Hi.

First, I must apologize but I don't have a background in statistical
analysis. I am reading a paper that is using two models of ordered mutiple
choice, logit and probit. I know that SPSS can build logit and probit
models, but are they ordered mutiple choice models? Is there a difference or
are all logit/probit models actually ordered multiple choice models.

Thanks.

Matthew

 
 
 

Logit/Probit Models

Post by Rich Ulric » Wed, 23 Jul 2003 23:52:12


On Mon, 21 Jul 2003 15:01:30 -0400, "Matthew"


> Hi.

> First, I must apologize but I don't have a background in statistical
> analysis. I am reading a paper that is using two models of ordered mutiple
> choice, logit and probit. I know that SPSS can build logit and probit
> models, but are they ordered mutiple choice models? Is there a difference or
> are all logit/probit models actually ordered multiple choice models.

Logit and probit represent continuous distributions.
Their first widespread use, so far as I know, was in
bioassay, where they were used to represent the proportions
of insects (say)  that died when given doses of poisons,
or the proportions of mice that were protected by vaccine.
That is a bit different from multiple-choice.
I have thought that "growth"  is what the models
are most often representing.

Assuming that your data come in that form,
what do you want to do with ordered multiple categories?

Well, you can *assume*  that they came from some
process (of growth;  of contagion; whatever)  that has
an underlying distribution - and use the corresponding
analysis.  The logit (assigning log of P/Q as a criterion)  
arises from a wide number of circumstances.  
The probit (assigning the Normal z  for a proportion)
arises in certain important instances;  and it became
the more popular  choice  50 years ago, I think, because
folks were already generally accustomed to the "normal"
distribution.

Hope this helps.

--

http://www.pitt.edu/~wpilib/index.html
"Taxes are the price we pay for civilization."  Justice Holmes.

 
 
 

Logit/Probit Models

Post by Arthur J. Kendal » Thu, 24 Jul 2003 03:20:09


I'm not sure what you mean by "ordered multiple choice".  I an
accustomed to hearing "multiple choice" used in description of the items
in a test.

SPSS has PLUM  which does polychotomous logistic models.  That is, the
dependent variable has more than 2 values, is categorical, and the
categories might be ordered.

Hope this helps.

Art

Social Research Consultants
University Park, MD USA
(301) 864-5570


> Hi.

> First, I must apologize but I don't have a background in statistical
> analysis. I am reading a paper that is using two models of ordered mutiple
> choice, logit and probit. I know that SPSS can build logit and probit
> models, but are they ordered mutiple choice models? Is there a difference or
> are all logit/probit models actually ordered multiple choice models.

> Thanks.

> Matthew