Page 94 - Design for Six Sigma for Service (Six SIGMA Operational Methods)

Customer Survey Design, Administration, and Analysis 75

In all these situations you need to determine how many missing, foreign, and
duplicated elements are in the sampling frame and how big a proportion
these wrong elements are as a percentage of the whole group of sampling
frame elements. If this proportion is large and it will affect the accuracy of
the poll, you should consider the possibility of using a different sampling
frame. For example, as stated before, in some of the opinion polls of the U.S.
2004 election, people with cell phones only were excluded in the opinion
poll sampling frame. If the portion of people excluded was a sizeable portion
of voters and their opinions were significantly different than those of tra-
ditional phone users, then this opinion poll might be rather unreliable.

4.4.2 Sampling Methods

Probability Sampling versus Nonprobability Sampling
Sampling methods can be classified into probability sampling and nonprob-
ability sampling. Probability sampling is used when you would like to draw
conclusions on the whole population based on the data you collected in the
sample. If your goal is just to learn something about the sample and you do
not intend to draw conclusions on the whole population, then probability
sampling is not necessary.

There are two characteristics of probability sampling:
1. The probability of selection is equal for all elements of the sampling
frame at all stages of the sampling process.
2. The selection of one element from the sampling frame is independent
of the selection of any other element.
For example, consider a sampling frame of 1000 people whose names
are written on equal-sized pieces of paper and where these paper pieces
are thoroughly mixed and selected one by one without the names on the
paper pieces being seen. If we assume that 100 people will be selected
for this sample, then the probability of selecting any person in the first
draw is 1/1000; the selection of any person in the second draw is 1/999;
… ; and selection of any person in the 100th draw is 1/901. Though the
probability of selecting a particular person is slightly different in each
draw, within each draw, the probability of selection for all the available
people is the same; this is consistent with the first rule of probability
sampling: The probability of selection is equal for all elements of the
sampling frame at all stages of the sampling process. Also, the prob-
ability of selecting a particular person is clearly independent of previous
drawings of other people, so this sampling practice is an example of
probability sampling.