Good question Carolyn...sometimes I lose my mind and slip into
biostat/epi-speak (used to teach Biostatistics to grad students).  Anyway,
"Ecologic Fallacy" is the error that occurs by assuming that because two or
more characteristics expressed as group indicies occur together, they are
therefore associated. For example, because a majority of a population has a
particular characteristic and a majority of the population also has the
disease, it may be assumed that the characteristic is associated with the
disease.  It is equally possible that the members of the population without
the characteristic are the ones who have the disease.  For those who should
know better statistically, we sometimes call this one method of lying with
statistics.  The way around this is to create incidence rates or relative
risk ratios for the specific subpopulations that you're examining.  In the
case of Prescott's research (while I like his conclusions), he examines
entire culture's patterns of parenting and bonding, and makes
generalizations about causality toward violence.  This may in fact be true,
but when looking at individual behaviors causing something, a researcher
needs to make sure the "unit of analysis" is the individual--NOT the entire
nation or culutre.

Is this clear?  Let me know if it's not.  I can give you more examples.
WarmLLLy,
Chris

: )   :  )   : )  : )   :  )   : )  : )   :  )   : )  : )   :  )   : )  : )

: )Chris Hafner-Eaton, PhD, MPH, CHES, IBCLC  email: [log in to unmask]  : )
: )HSR & Health Educational Consultant        voice/fax: 541 753 7340   : )
: )             **CHANGE THE WORLD, NURTURE A CHILD!**
: )   :  )   : )  : )   :  )   : )  : )   :  )   : )  : )   :  )   : )  : )