Sampling -- Systematic Sample
Return to the Sampling page
A "Systematic Sample" has the characteristic that there
is a predictable pattern to the selection process.
For example, to get a 60 student sample of the 11,413 credit students at
the community college in the fall term, we could just a list of those students and
then select the 190th student on the list, the 380th student on the list, the 570th student
on the list and so on until we select the 11,400th student on the list.
Using this methodology we do not need to consult a table of random numbers nor do we
need to use a random number generator. It may seem that we will have a representative
sample of all of the students, and, indeed we may. However, this procedure
is open to manipulation and unforeseen biases. Since we do not know how the original
list was generated, it is possible that the source of that generated list
intentionally or even unintentionally "stacked the deck" in making that list.
Certainly, if the method is known ahead of time the list could be arranged so
that we select just students with the highest current GPA. Even if the method is
not known, the list could have all of the students in alphabetic order. In that case,
selecting a student by the name of Palay means that a student
by the name of Parsons will not be chosen
since there are not 189 students with alphabetic names between the two.
This may not be a "problem" in that we may still get a representative sample, but
it is not equivalent to a simple random sample.
In a manufacturing setting it is
common to take samples of products for quality assurance tests.
A systematic approach to doing this might be to take the first product produced every hour.
In this way we would get a constant stream of products to be tested and we
would see products from early in the morning to closing time. However,
again, such a system would be open to manipulation. Machine operators, knowing that the
selection time is approaching may take more, or less, care in producing the next product or two.
Furthermore, some other "timed" event that might affect the quality
of the product might have its impact on our products
be over or under represented in such a sampling. For example,
regularly scheduled trains, passing at 8:20, 11:45, 2:10, and 4:30, might shake
our plant enough to affect the manufacturing process, and yet since we are not sampling
products made at those times, our systematic methodology will not detect any such effect.
Return to the Sampling page
©Roger M. Palay
Saline, MI 48176 December, 2015