From: rshepard@salmo.appl-ecosys.com   
      
   On 2003-12-24, Lionel Cox  wrote:   
      
   > My thesis is that fuzzy-sets are more accurate in the manipulation of data   
   > collected in the field of social science. In modelling terms by changing   
   > just one of the fuzzy-sets, according to one theory or another, overall   
   > change or differences can be displayed.   
      
   Lionel,   
      
     You've certainly explained yourself fully and well. I'll give you my   
   thoughts, and I'm sure that Bill, Earl or others will quickly correct any   
   mis-statements I write. For background, I am a field biologist (ecologist;   
   primarily a river ecologist/fluvial geomorphologist) and not a sociologist.   
   But, many of my clients will certify that I am a psychologist because of my   
   understanding of the mind of the regulator (small, so easy to understand :-)   
   ).   
      
     I would not describe fuzzy sets as useful for accurately manipulating   
   social science data. They permit us to quantify subjective concepts what are   
   easily expressed in language but not directly measurable. Your example of   
   'working class' is just such a concept. Furthermore, 'working_class' would   
   be a consequent fuzzy set, not an antecedent (one for which you directly   
   enter membership values). However, fuzzy sets can be manipulated with   
   mathematical rigor both directly and indirectly with approximate reasoning   
   expressed as IF-THEN rules.   
      
   > For example 'working class' is a relative term yet is used extensively in   
   > many social theories.   
      
     And this term is sufficiently vague, inherently imprecise and not directly   
   measurable to be a perfect application of fuzzy logic. (FWIW, almost all   
   applied environemntal science -- yes, that's redundant -- is also subjective   
   and based almost exclusively on societal values rather than quantitative   
   measurements.)   
      
   > If one was to create fuzzy-sets for example income, educational   
   > achievement, job and parental background (some of the properties of class   
   > recognition) I would anticipate that by using linguistic inputs with the   
   > results and displayed using fuzzy outputs they would differ dramatically   
   > to those that are calculated by official or government sources.   
      
     Agreed. Easily done -- using your questionaires or meetings with people to   
   gather values of those liguistic variables relative to included terms   
   (Income: none, low, moderate, high, CEO; Education: elementary, secondary,   
   tertiary, graduate).   
      
   > I suggest a persons definition of working class would vary by location and   
   > socialisation, London and the Shetland Isles, the UK & the USA for example.   
   > Working class in a city area being more marked as a label than in a more   
   > rural area. Hence the elicited information would give for example very   
   > working class, working class, maybe working class or not working class.   
   > The possibility is then to use the working class fuzzy-set in other   
   > question, say HealthCare provision where one of it properties would be   
   > working class.   
      
     Let me assume from the above that the question you want to answer is: does   
   the prevailing definition of "working class" vary with location? If so, what   
   attributes of location cause the observed differences in definition?   
      
   > So your answer being 'yes' where do I find such a questionnaire that has   
   > been published along the lines I describe.   
      
     Nothing done by someone else will suit your needs. If the purpose of the   
   questionnaire is to elicit values and the membership function shape for each   
   linguistic variable and fuzzy set you deem important, then write your   
   questionnaire to give you usable answers.   
      
     You need to start at the end and work your way back. If my above assumption   
   is correct, then your questionnaire needs a section describing the   
   respondent by all the characteristics you belive pertinent to varying   
   definitions of "working class". Then it needs a section for each variable,   
   and the questions need to be carefully worded to provide the information you   
   need while not being so onerous that the potential respondent doesn't.   
      
     This is why you will not find your questionnaire already prepared for you   
   on some Web site. But, you can certainly do the job. And, when you're done   
   thinking this through in detail, you might find that an Ordered Weighted   
   Average (OWA; look for publications by Ron Yager) is a way of interpreting   
   the relative importances of the various factors contributing to the concept   
   of "working class".   
      
   Have fun!   
      
   Rich   
      
   --   
   Dr. Richard B. Shepard, President   
   Applied Ecosystem Services, Inc. (TM)   
      
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)