In its Feasibility Study (see table below), Stat Can illustrated the sample sizes necessary to achieve the precision previously described, for estimates pertaining to individuals. The cv's illustrated correspond to a min p of 0.10. It is assumed that this precision is needed at the region level, for the Inuit population only. This requirement is indicated by the shaded cells. For example, a sample size of 677 Inuit people is necessary in Nunavik in order to obtain a cv of 16.5% for a proportion of 0.10 of the Inuit population.
The last two columns of the table show what the total sample size (Inuit persons plus non-Inuit persons) would be if the proposed design of collecting data for both Inuit and non-Inuit people is chosen, and the cv's of the corresponding estimates. Resulting samples and cv's for the three sub-regions of Nunavut are also shown.
All sample sizes in Stat Can's table reproduced below are called basic since they are the minimum sizes needed to meet the requirements at the region level.
Table A: Basic sample sizes by region
due to the low
proportion of Inuit in the
communities, they have been excluded from the sample size calculations for
Labrador: Port Hope Simpson;
Cartwright; Division No. 10, Subdivision C; Happy Valley-Goose Bay; Labrador
City; and Division No. 10, Subdivision E.
The variation in sample frames and regional populations across SLiCA countries means that the above table is only illustrative of the precise sample sizes needed. However, the above illustration was used by the SLiCA team as the basis for a target sample size for each reporting region of approximately 600.
National Differences in Sample Size:
Subsequent to this report, Statistics Canada agreed to a proposal by the SLiCA indigenous Steering Committee to combine SLiCA in Canada with Stat Can's Aboriginal People's Survey (APS). APS was designed to provide reliable results (for selected variables) at the community level. Therefore the required sample sizes were much larger in Canada than originally envisioned for SLiCA where regional level results are targeted.
For the on-community portion of 2001 APS (which included the 53 Inuit communities in arctic regions), a targeted minimum proportion of 10% with a maximum CV of 25% were fixed for the adults. The strategy was then to estimate the number of adults required in the sample in order to meet these requirements and apply the same sampling fraction to the children. Note that because the number of children is smaller than the number of adults for most, if not all communities, less precision will be achieved for the children. This means that, for a characteristic present for 10% of the children, the CV will be larger than 25% (less precise). Alternatively, for a CV of 25%, the minimum proportion of children having a specific characteristic will have to be greater than 10%. Note that these calculations were made by community rather than by region, as in the case of the initial Feasibility Study. Therefore the sample sizes required are much larger.
The sample size also considered expected levels of response. Projected response rates ranged from 75% on First Nation reserves to 90% on Inuit communities. These were estimated using what had been obtained in 1991 APS. Since observed response rates were often larger then what had been anticipated, this could lead to smaller CV's in the estimation phase in certain communities.
The total sample size for the 53 Inuit communities combined was 10,775. The following communities were included:
Alaska's originally proposed design called for each of the Iņupiat sub-regions to be considered a reporting region. Based on Stat Can's Feasibility Study, this would require samples of approximately 600 in each region. Based on funding constraints, we changed our design to consider the three Iņupiat sub-regions combined as a primary SLiCA reporting region. We therefore designed the sample to yield approximately 600 interviews.
While 600 interviews is required to meet Stat Can's original level of reliability, lower sample sizes can nevertheless provide unbiased estimates for sub-populations at a lower level of reliability. We decided that it was important to provide regional Native organizations with separate tabulations and therefore targeted 200 interviews in each of the three Iņupiat sub-regions: North Slope, Northwest Arctic, and Bering Straits.
Applying the sample methods described earlier, the following shows the samples selected for Alaska communities:
The Swedish Saami sample was drawn with a random start and sampling interval from a list ordered by community. There was only one stage of sampling (individuals). Therefore the sample is highly dispersed geographically in proportion to the actual distribution of the Sameting voting list.
Applying the sample design described earlier with adjustments for anticipated non-indigenous households and non-completed interviews, the following is the selected sample:
Applying Greenland's sample design, the following are targeted numbers of respondents by city, settlement, and region:
Please continue with Interview Methods.