Sampling and Weighting Technical Report, Census of Population, 2016
2. SamplingNote 1

 

When a sample survey is conducted, the sample selection must be planned properly. In sampling, a subset of the survey's target population is selected to receive the questionnaire. The responses of the subset are used to draw inferences for the entire population. Two types of sampling exist: probability sampling and non-probability sampling. Probability sampling is preferable when producing statistical inferences for the entire population is important, since the probability of unit selection can be calculated and the sampling error can be estimated. This chapter discusses the selection of the sample that received the 2016 Census long-form questionnaire.

2.1 Long-form sample universe

The census household universe was broken down into three parts: private households, collective households and households outside Canada. The long-form sample universe consists only of private households, including those living in private dwellings attached to collective dwellings in Canada. This universe excludes incompletely enumerated Indian reserves and settlements. Unless otherwise specified, the term "in scope" indicates that a household is part of the long-form sample universe (i.e., private households that are not living in incompletely enumerated Indian reserves and settlements). "Out of scope" refers to households not in the universe (i.e., households living in collective dwellings, outside Canada, or in incompletely enumerated Indian reserves and settlements).

2.2 Long-form sampling design

In most cases, the long-form questionnaire was distributed to one-quarter of the households in the long-form universe to gather demographic and socioeconomic data on the Canadian population. The sample was selected from the list of dwellings for the 2016 Census of Population. At the time the sample was selected, the addresses of out-of-scope dwellings were unknown. This meant that some dwellings erroneously received a long-form questionnaire. Once a dwelling was determined to be out of scope, no further collection or processing activities were carried out.

Dwellings were selected to receive the long-form questionnaire according to a stratified systematic sampling design. The sampling design strata were the CUs. For mail-out CUs, the sampling was systematic, with a one-quarter sampling fraction. The selection starting point was random. For list/leave CUs, sampling was systematic, and the sampled dwellings were every fourth one on the list, i.e., 4th, 8th, 12th, etc. For example, if a list/leave CU had seven dwellings, only one dwelling was selected. Finally, in canvasser CUs, all households were selected. These CUs were take-all strata.

The sampling design had one exception. Private dwellings attached to collective dwellings were added to the sample with certainty. However, they completed only the short-form questionnaire. Long-form questionnaire responses were later imputed for these households.

Except private households attached to collective dwellings, all households selected for the sample were asked to complete the long-form census questionnaire. Households in private dwellings that were not part of the long-form sample were asked to fill out the short-form questionnaire.

Note

Date modified: