Sometimes I get asked how the "20% sample" for the results from the 2B Form of the Census can be construed as answers for the "whole" (ie. 100%) Canadian population.
Would anyone care to give me a phrase or two about how they would explain that the results of the "20% sample" can be trusted as an extrapolation for Canada as a whole?
1) Depending how technical of an answer you wish to provide, the 2001 Census Handbook has a straightforward explanation for the use of the 20 percent sample in the Census on page 3. For more technical explanations of the sample and coverage in the Census, check out the technical reports, 92-395-XIE (about sampling and weighting) and 92-394-XIE (about coverage). Section 4 of 92-395-XIE addresses the estimation methodology used in the Census. It may be useful to read Section 2 in this same report first because it describes the data processing used with the Census. There are a lot of techniques employed in providing Canadians with the final Census results and this section might help put in perspective the overall process used.
2) Here is a brief description of weighting from a census FAQ we have...
Many of the census questions are asked only for a sample of 1 in 5 households. In order to make the results for the sample refer to the total population, the answers to these questions for each household in the sample are multiplied by a weight. The weight for each household depends on the exact characteristics of the household, but is generally around 5.
For an in-depth explanation on sampling and weighting in the census please refer to the technical report: