Thirty jackknife replicate weights were developed for use in determining the standard errors of the survey estimates. The 30 replicates cut across strata – 30 replicates were created within each province by sample type and by stratum.
In the urban strata, households were selected systematically in the first stage of sample selection. The replicates were formed by sorting the households in the order that was used for the systematic sample selection, and by assigning replicate numbers sequentially from 1 to 30 to the households, restarting back to 1 after reaching 30.
In the rural strata, PSUs were selected in the first stage of selection. Since fewer than 30 PSUs were selected in all strata, the PSUs were split to form the replicates. As much as possible, the PSUs were split into an equal number of replicates. If this was not possible, the PSUs with a larger number of respondents were split into more replicates. For example, if 11 PSUs were selected for a particular province and sample, then the 8 PSUs with the greatest number of respondents would each be split into 3 replicates, and the remaining 3 PSUs would each be split into 2 replicates, to obtain a total of 30 replicates.
The replicates were formed using the initial IALSS sample of over 40,000 units. Initial jackknife weights were calculated, based on the design weights. The entire weighting process was repeated for each of the 30 jackknife weights, including non-response weighting adjustments, integration of the weights, and calibration.
OECD and Statistics Canada (2000), Literacy in the information age: Final results of the International Adult Literacy Survey. Paris and Ottawa: Authors.