Nick,
Doesn't sound too tricky - as you describe it, it seems a pretty good
candidate for some form of Bayesian analysis: p(A|B) is proportional to
p(B|A)p(A), where B="is described as a pansy" and A="is actually a pansy"
You can probably get good empirical values for your priors ("what IS the
probability of finding a pansy round here?"), which is a pleasant change for
Bayesians as they usually guess these values and then spend time trying to
convince you that the final result isn't sensitive to the priors anyway.
Also I'd expect that you can probably get reasonable values for
those conditional probabilities, in consultation with your local flower
expert.
Robert
On 2/15/08, Nicholas Thompson <[EMAIL PROTECTED]> wrote:
>
> All --
>
> Has anybody thought about how to make use of truly lousy data? There are
> increasingly sources of public data on subject matters such as weather and
> (see below) flowers and birds where the quality of the data is truly awful
> by ordinary standards and yet there is so much of it that it seems a crime
> not to try to make use of it. So Sally writes in to say that her morning
> glories are in bloom in April when what she means is her pansies. Her
> neighbor gets the pansies right but screws up on the tithonia. Is there
> any way to add this all up and get something?
>
> thoughts?
>
> nick
>
>
>
>
>
> Nicholas S. Thompson
> Research Associate, Redfish Group, Santa Fe, NM ([EMAIL PROTECTED])
> Professor of Psychology and Ethology, Clark University
> ([EMAIL PROTECTED])
>
>
============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org
