Transcript An Analysis of Person Duplication in Census 2000

Discussion: Play-Acting at
Science?
Robert E. Fay
U.S. Census Bureau
1
Disclaimer
My written discussion advocates specific
directions for survey research.
Disclaimer: My discussion represents my views
and not necessarily those of the Census
Bureau.
General points
My written comments are general and not
specific to Saris’ talk
I argue against quantified total error models at
this point (and possibly ever)
Part of my criticism stems from my study of
scientific communication and its importance
to hard science
The past
I comment on the total error modeling of the
decennial census undercount in the 1990
and 2000 census, based on its failure of
scientific communication
Future directions
I encourage more active collaboration of
statisticians and behavioral scientists –
example: mathematical biology
APS as a model for some of the behavioral
scientists I want to involve
Future directions
I encourage what I call more “fine-grained”
analysis. I provide two examples published
in Science. Some talks presented today
exemplify that direction.
I would like to see the reporting of research
results more closely modeled on the hard
sciences
Written remarks
Note: These transparencies touch on a few
points from more extensive written remarks.
You may contact me for the written version.
[email protected]