Appendix II: Margins of Error
Contents |
>> Next: Appendix III: References
Precision in general
Because we have only taken a sample of New Zealand adults, any results represented for this population will have a margin of error.
Two issues need be considered with respect to precision of results and margins of error for segmentation results like these: the judgemental component of segmentation, and weighting.
The judgemental component of segmentation. The large amount of judgement rather than statistics involved in the segmentation process means that margins of error are not calculable for the fundamental split into segments (and relatedly, the size of each segment). For example, no clear statistical criterion prevented us from choosing a split into four segments rather than six or from clustering 20 key variables rather than 13.
Weighting. Results are weighted to correct for the probability of selection and sample imbalances (e.g. age, gender, ethnicity) as described in detail in the Technical Report. Such weighting typically results in margins of error distinctly larger than those for a simple random sample of the same size.
Indicative margins of error
The indicative margins of error provided below are those that would apply for a simple random sample of the “effective sample size” shown. As described in the weighting section of the Technical Report, the effective sample size (for the full sample) is approximately half the actual (unweighted) sample size.
This does not take sample stratification or weighting non-linearities into account, but these are not expected to have had a major effect.
For the rim-weighting procedure used, more precise margin of error calculations would be time-consuming and hence expensive; incurring these extra costs was not seen as worthwhile given that the focus of the study is on segmentation rather than, for example, prevalence estimates.
The reports commonly compare results from a segment to the Target Group overall. The margin of error for such differences is slightly larger again (because of the additional sampling error in the Target Group estimate).
For example, an indicative margin of error for difference between the Others Oriented segment versus others in the Target Group (n=3685–398=3287) where both estimates are around 50% is 7.4% rather than the 6.9% in the table above (for a standalone estimate from the Others Oriented segment).
Table 1: Indicative margins of error
Actual sample size
Effective sample size (approx.) |
Others Oriented
398
199 |
Support Seekers
529
265 |
Target Group
3685
1843 |
Estimate in report
50%
40% or 60%
30% or 70%
20% or 80%
10% or 90%
5% or 95% |
6.9%
6.8%
6.4%
5.6%
4.2%
3.0%
|
6.0%
5.9%
5.5%
4.8%
3.6%
2.6% |
2.3%
2.2%
2.1%
1.8%
1.4%
1.0% |
Note: 95% confidence level used.
Contents | >> Next: Appendix III: References
Updated | 22 Mar 2007.
Skip page tools.
