Department of Statistics Seminar
North Carolina State University
presents
Dr. Chuck Proctor
North Carolina State University
"The Underworld of Statistics behind Bulk Sampling with a Composite"
ABSTRACT
A whole lot of sampling of
materials and of the environment goes on without benefit of random
numbers, but it could be done with their aid. This is due in large
part to the lack of a sampling frame, which is a list of every
place-time where an observation could be made or, in the case of a
lot of coal, where some material could be brought into the sample.
When a frame is actually constructed and used, it turns out that it
does not lead to a partitioning of the lot. It does, however, create
a finite but very long list of possible outcomes (derived from all
the ways the random numbers could have arisen) and this list deserves
to be recognized as the sampled population. Recognition of this population of
cross-sectional test portions, as they are called when making an ash
determination, caused us to reformulate the variance formula for bulk
sampling. Sampling variance formulas come in two varieties. One is
used to estimate the standard error of a survey estimate based on
data from that same survey, e.g., s2/n. The other
is used to design a sampling plan and is based on population
parameters, e.g.,
/n.
The formula we derived is of the second type and is:
IS(1+CV12)(1+CV22)(1+CV32),
IS is
increment selection variance and depends on the number, sizes and
spacings of the increments, while the CV's refer to how well the
evenness of the increments is preserved through the stages of sample
preparation and analysis.
In the talk we will first review this frame construction step and then derive the formula. We will then go back to review the literature, so to speak. Three themes seem to have guided statistical theories of bulk sampling, (1) general statistical theory, including time series, (2) particles as sampling units, and (3) enumerative survey sampling theory. We will attempt to show that enumerative sampling theory without frames or with only hypothetical frames may have retarded development of theory adequate for understanding the statistical consequences of the compositing operation.
Friday, January, 11, 2002
3:35 - 4:35 pm
206 Cox Hall
Refreshments will be served on the second floor of Dabney Hall (left of Room 222) at 3:00 pm.