and the sample is selected without replacement. The higher the sampling fraction (n/N), the lower the fpc and
the standard error of estimates based on the sample.
Guideline 7.15.Design effect. If probability cluster sampling is used, the targeted sample size should
be adjusted taking into account the design effect.
The formulas presented in the above discussion of the confidence interval and the hypothesis testing
approaches in determining the size of a sample assume that simple random sampling will be used. On the
other hand, other formulas must be used for alternative sample designs. A review of these formulas is beyond
the scope of this text. Yet, an adjustment may be made via the targeted sample size by applying the design
effect. The design effect (DEFF) is the ratio of the variances of sample design employed to the variances of
a comparable simple random sample design. The DEFF of a stratified sample design tends to be a little less
than one, indicating that if stratification is used the sample size may be smaller than the sample size simple
random sampling at the same margin of error. Technically, the DEFF indicates how much less (or more)
the precision of a nonsimple random design used when it is compared to the precision of simple random
sample design. From a sample size perspective, it indicates how many more (or fewer) elements should be
selected in the planned sample design compared to the sample size required for a simple random sample to
achieve the same level of sampling variance. If the DEFF of a cluster sample is greater than 2 (a DEFF of
2.0 is typically a default value), the sample size for the sample must be more than twice the sample size of a
comparable simple random sample at the same margin of error.
Guideline 7.16.Attrition/mortality rate. The targeted sample size should be adjusted to take into
account the attrition or mortality rate.
If a longitudinal study is planned, in particular a panel study, attrition should be anticipated. The initial sample
size should be adjusted to take this factor into account.
Summary
The choice of sample size is a very important decision. Guidelines for choosing the size of a sample indicate
that such factors as having an exploratory research objective, the minimization of the burden on study
participants, homogeneous population, scattered population, and limited resources suggest a smaller sample
size rather than a larger sample size. On the other hand, such factors as quantitative, nonexperimental, and
longitudinal research designs and a complex and detailed data analysis design suggest a larger sample size
rather than a smaller sample size. “Rules of thumb” are suggested for nonprobability sample designs, and
statistical formulas are suggested for probability sample designs. The statistical formulas take into account
such factors as confidence intervals, level of significance, level of power, and effect size. The final sample
size should be calculated after making adjustments for the incidence rate, the nonresponse rate, the finite
population correction factor, the design effect, and the attrition/mortality rate.
SAGE
2012 SAGE Publications, Ltd. All Rights Reserved.
SAGE Research Methods
Page 15 of 17
Sampling Essentials: Practical Guidelines for Making Sampling Choices