Table 4
Summary of recommendations for users of the Canadian Health Measures Survey relationship data files for conducting paired respondent analysis
| Recommendation number | Recommendation topic area/subject/aspect | Recommendation |
|---|---|---|
| 1 | Relationship types | Birth parent–child pairs are appropriate for analysis. Sibling pairs have limited analytical value. Analyses using any other relationship type are not recommended with five (or fewer) cycles of data. |
| 2 | Age groups of the child | Data for children aged 3 to 11 years are available for analysis. The two age groups—preschoolers (aged 3 to 5 years; cycle 2 onward) and 6- to 11-year-olds (cycle 1 onward)—can be examined together or separately. |
| 3 | Sample sizes | For reliable estimates, approximately 500 or more pairs per analytical subgroup are recommended; however, this depends on the analysis and outcomes of interest. Lower prevalence estimates will require larger sample sizes. Data users are encouraged to combine as many cycles as possible when conducting paired analyses to maximize sample size. |
| 4 | Survey weights | The survey weight variable depends on the unit of analysis. If the outcome of interest is for the child, use the child-assigned weight. If the outcome of interest is for the parent, use the parent-assigned weight. |
| 5 | Bias | Despite the potential biases related to household composition and specific second member selection or inclusion, children whose second household member is a respondent birth parent are sufficiently representative of children in Canada. However, data users should be aware of the biases when interpreting and discussing their results. |
| 6 | Subsample files | In addition to the full sample data, use of the accelerometer subsample file is analytically possible for birth parent–child pairs when three or more cycles are pooled. Analyses of sibling pairs are possible for 6- to 11-year-olds only, using all five cycles of data. The fasting subsample is possible among 6- to 11-year-old children with birth parent–child pairs only, and all five cycles of data should be pooled. |
| Source: Langlois K, Bushnik T, Colley RC, et al. Using the Relationship Files and Paired Respondent Data in the Canadian Health Measures Survey: Feasibility Study—An Update. Statistics Canada internal document available upon request. February 2020. | ||
- Date modified: