Date: 	Mon, 23 Feb 1998 16:51:48 -0500 (EST)
From: Rakesh Bhandari <bhandari-AT-phoenix.Princeton.EDU>
Subject: RE: BHA: Need Help Fast


Nick wrote:
> proposal.  My thoughts on the discussion re statistical methods are that
> perhaps non-inferential statistics can play a part in a quantitative CR
> approach.  I have in mind techniques like multi-dimensional scaling and
> cluster analysis...  
>   They make no distributional assumptions which would run
> counter to the idea that closure doesn't exist in the social world.
> Secondly, and more interestingly, they aim at identifying *similarities*
> between and within data rather than with the amount of variation in a
> dependent variable contingent on the behaviour of independents.  
I know this is an idiosyncratic idea, but looking over a discussion of 
Bourdieu's  correspondence analysis of cross-tabulated data, I get the 
sense that such graphical presentations are most useful for advertisers so
they can have a profile of target sub-populations. As I have said before,
I thought the questions raised by Andrew Sayer about regression techniques
are profound.
Best,
Rakesh




     --- from list bhaskar-AT-lists.village.virginia.edu ---