Skip to content

Continuous (Non-Count), Skewed Data With Many Zeros

2 messages · Highland Statistics Ltd, Rich Shepard

#
On 16/05/2012 06:00, r-sig-ecology-request at r-project.org wrote:
There is a limit to what can be done with stats. Do you really want to 
analyse data that has been collected by different folks, labs, etc.? Any 
change over time may actually reflect a lab effect.

If you have a detection limit then you may want to consider models with 
a truncated distribution..:-)

Alain
#
On Wed, 16 May 2012, Highland Statistics Ltd wrote:

            
Alain,

   Within the ivory tower of academic ecology such considerations have
weight. The world of the environmental regulator is a totally different
situation. Yes, laboratories and methods have changed over the decades, and
detection limits have lowered in some cases. But, these are the data used to
issue or renew operating permits and businesses need those permits to
survive and prosper.
Yep. This is one option I'm going to look at very closely. I started
reading the pdf accompanying the NADA package and have ordered the book on
which it's based.

Regards,

Rich