Re: st: NBREG for ordinal scales

[David Bell <dcbell@iupui.edu>]

On Oct 10, 2006, at 1:08 PM, n j cox wrote:

> I think it's pretty much wired in that Poisson,
> negative binomial, etc., really are for counts.
Actually, I seem to recall that Poisson processes are based on  
probabilities of changes of state.  If you are counting the  
sequential changes, then you have count data and it is  
straightforward to label each state with the number of persons (or  
other entities) the state represents.  If, on the other hand, the  
changes are psychological (such as the change from being strongly  
opposed to some action to being "only opposed") then the labels for  
the states are not counts.  As I recall, James Coleman, in  
Introduction to Mathematical Sociology (1965) used Poisson models of  
responses to ordinal attitudinal scales.

The approach never became popular in sociology, but it gives a  
justification for using Poisson and related processes on non-count  
data, as Matthew seems to want to do.

Dave
====================================
David C. Bell
Sociology
Indiana University Purdue University Indianapolis (IUPUI)
====================================


*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/