Design Effects


 
The design effect, DEFF, is defined as the ratio of the true variance of a statistic under the actual design divided by the variance that would have been obtained from a simple random sample of the same size(1).

   DEFF(u^) =  Var(u^)
                         SRS Var(u^) 

  • where SRS Var(u) is the variance that would have resulted under simple random sampling assumptions(1)
  • Var(u) is the variance of the statistic under the actual design
Why is design effect important?
.
Efficient sample size calculations are based on an estimate of the sample size required to limit sampling variability to the desired level.  Efficient sample size calculations assume simple random samples(2). Therefore, sample designs other than simple random sampling have an impact, called design effects, on sampling variability.  As a result of this impact, design effects are important considerations when determining sample size. 
     The design effect represents the cumulative effect of  design components such as stratification, unequal weighting, and clustering, and will differ for each design(1). For example, sampling variability increases when cluster sampling is used rather than simple random sampling(2).
How is DEFF used?

The design effect is a direct way of addressing the impact of design on sampling variability.
The design effect can be multiplied by the expected sampling variance in the calculation of an efficient sample size to adjust for the impact of the design. In order to incorporate the effect  of the design into the calculation of the efficient sample size, the researcher must estimate the design effect and this is usually based on past survey experience as statistical literature provides little guidance(1).

References

1.     Rosander AC. Case Studies in Sample Design. New York: M. Dekker, 1977.

2.     Henry GT. Practical Sampling. Newbury Park: Sage Publications, 1990.

HOME