Home
Up

Sample Frames

The term "sample frame" is defined as the list or lists from which individuals or households are selected. A good sample frame:

Includes all individuals in the target population
Excludes all individuals not in the target population
Includes accurate information that can be used to contact selected individuals

National differences in statistical reporting mean that we have to use different types of sample frames in each country. To better understand these differences, see a working memo, "SLICA Sampling Strategies".

Here are separate sample frame overviews by region:

Canada
Alaska
Sweden
Chukotka
Greenland

Canada

The sample frame for Canada is based on responses to six questions in the 2001 Canada Census. Based on these responses, individuals who were reported by the household member completing the form to be themselves or their ancestors part an aboriginal group were included in the Aboriginal Public Survey (APS) sample frame. The APS included an Arctic special section derived from the SLiCA international core questionnaire. Additional SLiCA core questions were contained in the general APS. SLiCA comparisons will comprise a subset of the APS sample frame, those individuals residing in private dwellings in the Inuvialuit, Nunavik, Nunavut, and Labrador Inuit regions of Canada. For more detail, see Canada APS Concepts and Methods Guide.

Alaska

The Alaska SLiCA team did not have access to U.S. census population lists as in the case of Canada. Instead, the sample frame in Alaska consisted of four components: regions (N. Slope, Northwest Arctic, Bering Straits), communities, blocks (in the regional centers only), housing units in all communities, and individuals in households.

Regional Sample Frame

We treated each Iñupiat settlement region as a separate sample frame.

Community Sample Frame

Within each region, we:

Selected the regional centers (Barrow, Kotzebue, and Nome) as self-representing.

In the Northwest Arctic and Bering Straits regions, we sorted communities by cultural/linguistic grouping and assigned measures of size corresponding to result of dividing their Native population aged 16 and older by 25. We then picked a random start between 1 and applied a sampling interval by measure of size (e.g. a sampling interval of 25, taking every 25th measure of size) to select five villages with each region. All villages on the North Slope were selected because there are only eight.

Block Sample Frame

We sampled blocks within Kotzebue and Nome. We used 2000 census counts of the number of indigenous householders in each block. We assigned a measure of size to each block. We selected blocks with the same random start and sampling interval procedure described for sampling communities. In Barrow, we sampled households directly.

Household Sample Frame

In Kotzebue and Nome we listed (i.e. mapped and wrote a description of each housing unit) households in sampled blocks. We used 2000 Census maps and each household's unique physical address number as the primary identifiers.

In Barrow and the sampled villages in each region, we listed all households. With the aid of aerial photographs modified to show all buildings, we worked with local residents to identify occupied households and to eliminate from the household list households known with certainty not to have at least one indigenous household member.

Individual Sample Frame

All persons 16 and over who considered themselves to be Alaska Native and resided in sampled households were part of the individual sample frame. We selected individuals by choosing the person who has the next birthday. This has been determined to for all practical purposes be a random variable.

Sweden

The sample frame originally selected for use in Sweden is a list ("Vilhelmina List") intended to include the descendants of all Saami identified in a past study (better cite will be provided). Current contact information for persons on this list is maintained separately by the Sweden Statistical Bureau. The Swedish SLiCA team belatedly discovered that contact information could not be obtained without prior permission of each selected respondent.

In view of the fact that interviewers had already been hired and interviewer training was about to begin, we explored options for alternative sample frames. The first option we explored was to start in communities where the concentration of Saami households is sufficiently high to use area probability sampling techniques (i.e. listing housing units and screening sampled units for Saami household members). We matched the Vilhelmina total population counts with the figures by Kommun. The results were not good: the highest concentration was 24% for Jokkmokk. That isn't high enough for area probability sampling to be feasible.

We then explored the option of merging multiple local lists of Saami to build a sample frame in the high concentration communities like Jokkmokk. On discussing this option with the local interviewers, we concluded that the local lists would not be sufficiently inclusive to be defensible as a total population sample frame for Saami. In addition, we concluded that they would probably only be available in hard copy form, meaning that we would face the labor-intensive task of entering the separate lists as databases prior to merging.

We then considered the Sameting voting list, understanding that this list is not an inclusive sample frame either. Our first question was how it compares to the Vilhelmina list. We successfully obtained the Sameting voting list in hard copy, both as a complete list and as a subset built from people with postal codes beginning with "9" and "8" and the part of "7" which together essentially constitutes the Sápmi area. Comparing total counts in these lists with the Vilhelmina list, we found that the Sameting list is 20% of the entire Vilhelmina list and 28% of the Vilhelmina list portion limited to the Sápmi area, which is the area of concern to us in this study.

We concluded the Sameting voting list is the best choice as a Phase 1 sample frame for the following reasons:

It is a definable subset of the Vilhelmina list, meaning we can sample from the Vilhelmina list later and have two probability samples that can be combined when properly weighted.
It likely includes the vast majority of herding-related households, a sub-group we already know we want to oversample
It is in itself a significant portion of the Vilhelmina list (28%)
Neither the alternatives of area probability sampling nor merging population lists are feasible.
We could immediate use the Sameting list (in hard copy) with names and addresses to use as a Phase 1 sample frame.

Chukotka

The Chukotka sample frame has five components: regions, districts, communities, households, and individuals.

Region Sample Frame

There are four regions: Anadyr, Central, Western, and Eastern. Each district is self-representing.

District Sample Frame

There are, including Anadyr, eight districts. Each district is self-representing.

Community Sample Frame

Within each region/district combination we grouped communities by three size classes. We used census counts of the indigenous population to assign measures of size to each community. We picked a random start and selected communities according to a sampling interval on measures of size. We selected a total of 24 communities.

Household Sample Frame

The households sample frame is derived from municipal household lists. Local residents will help to eliminate households certain to contain no indigenous household members.

Individual Sample Frame

All persons 16 and over who considered themselves to be indigenous and residing in sampled households are part of the individual sample frame. We select individuals by choosing the person who has the next birthday. This has been determined to for all practical purposes be a random variable.

Greenland

Greenland is able to use its administrative record system as the sample frame. Please see Greenland Sample Design.

Please continue with Sample Methods.