Page 97 - Design for Six Sigma for Service (Six SIGMA Operational Methods)

78 Chapter Four

relationship between the sample and population or use such a useful prob-
ability distribution model as the normal distribution. However, nonprob-
ability sampling can still be helpful. It is much easier to select a sample and
get a feel of what a portion of customers may think. For example, nonprob-
ability sampling can be used to quickly select a small sample of respondents
(say 30) to pretest a survey instrument. Although the conclusion from these
30 people cannot be generalized to the general population, a lot of short-
comings of the survey instrument can be identified.

There are a few commonly used nonprobability sampling methods. The
most commonly used sampling method is the sidewalk survey. The
interviewer, for instance, may interview passersby near a shopping center,
assuming the general population is the shoppers. In this approach, the
sampling frame is not explicitly identified and numbered; the probability of
selecting any particular passerby is unknown. The advantage of this approach
is the ability to get a lot of information quickly. The other commonly used
nonprobability sampling technique is snowball sampling. Snowball sampling
is particularly beneficial in instances where it is difficult to identify potential
respondents. Once a few respondents are identified and interviewed, they are
asked to identify others who might qualify as respondents. Soon the list of
respondents will be increased.

4.4.3 Sample Size Determination
One critical question in a survey project is how many units in a sample are
needed so that the analysis result derived by this sample can be generalized
to the whole population. The answer to this question depends on two key
factors. One key factor is what level of accuracy is required in this study; the
greater level of accuracy required in the study, the larger the sample size
needed. The other factor is the cost and time that we would like to spend in
this survey study; a larger sample size will certainly mean higher cost and
longer time. Therefore, the sample size is mostly determined by the tradeoff
between desired level of accuracy and cost and time.

Determination of Sample Size for Variables Expressed in Proportions
In survey data analysis, many variables are expressed in terms of proportions.
For example, we could ask customers:

Do you like the service of ABC Bank?
Yes _______ No _______
The proportion of people in the survey sample answering yes, which is often
ˆ p
called the sample proportion , is frequently used as the statistical estimate