Data quality, concepts and methodology

Data limitations

The figures appearing in this report are weighted estimates based on the data collected from the sample of wave two respondents and are subject to potential error. In all surveys, error can be divided into two components: sampling and non-sampling error.

Sampling error is the difference between an estimate derived from a sample and the one that would have been obtained based on information from everyone in the population using the same collection procedures. The size of the sampling error can be estimated from the survey results. Indications of the magnitude of the sampling error for the estimates must be provided. If the estimated sampling error is greater than 33.3% of the estimate, it is considered too unreliable to publish. Although not considered unreliable to publish, values with an estimated error between 16.6% and 33.3% of the related estimate should be used with caution. In this report, all values had sampling errors lower than 16.6% of the related estimate.

All other types of errors—such as coverage, response, processing and non-response—are considered non-sampling errors. Many of these errors are difficult to identify and quantify. Coverage errors arise when there are differences between the target population and the surveyed population.

Of the 12,040 immigrants followed for the wave two interview, 9,322 participated (respondents); 1,370 chose not to participate (non-respondents); and 200 were found to be no longer in the population of interest (out of scope). Additionally, 1,148 immigrants could not be located, and thus their status was unresolved.

To the extent that non-responding immigrants differ from immigrants who were unresolved and—more importantly—from immigrants who responded to the survey, separate weight adjustments were introduced in the final weights in order to compensate for the non-completed questionnaires.

While there were some partially completed questionnaires and non-response items, such omissions were in general infrequent. However non response was higher for questions on family and household income amounts. Hence imputation methods were used to correct these errors.

For more details about the LSIC sample design, use of weights, non-response adjustments and imputation, please refer to the Microdata User Guide – Longitudinal Survey of Immigrants to Canada – Wave 1 ¹ (a wave two user guide will be available in November 2005).