A personal opinion (not Cleveland's fault) is that a lot of the
apparent conflict between traditional inferential statistics and EDA
is due to sociological causes. I think you gather data and do EDA
until you know enough about the phenomonon to frame a sensible
hypothesis or model. Then you design an experiment or otherwise
gather NEW data to test your hypothesis or estimate the parameters of
your model. Of course, you use EDA techniques on the new data to make
sure your methods satisfy assumptions and your model is a reasonable
one. Each type of statistics has a role to play in this process --
they are not competitors nor mutually exclusive.
This sequential, interactive process takes years -- even generations.
In our society we have a peculiar institution called the thesis or
dissertation that forces people to try to cram all this into a year or
two. The result is lots of inference based on models or hypotheses
with no empirical justification. Indeed, sometimes there is no
justification of any kind, and sometimes the (are they perpetrators or
victims?) folks involved have no idea what the assumpions or the model
they are working with IS.
a quotation of a p-value is part of the ritual of
science [??], a sprinkling of the holy waters in an
effort to sanctify the data analysis and turn
consumers of the results into true believers.
Cleveland, ibid, p.177
_
| | Robert W. Hayden
| | Department of Mathematics
/ | Plymouth State College
| | Plymouth, New Hampshire 03264 USA
| * |
/ | (603) 968-9914
| ) (603) 535-2489
L_____/ hayden@oz.plymouth.edu