Steven Thomas1, Brenda Wannell. 1. Household Survey Methods Division, Statistics Canada, Ottawa, Ontario. Steven.Thomas@statcan.gc.ca
Abstract
BACKGROUND: A single cycle of the Canadian Community Health Survey (CCHS) may not meet researchers' analytical needs. This article presents methods of combining CCHS cycles and discusses issues to consider if these data are to be combined. An empirical example illustrates the proposed methods. DATA AND METHODS: Two methods can be used to combine CCHS cycles: the separate approach and the pooled approach. With the separate approach, estimates are calculated for each cycle separately and then combined. The pooled approach combines data at the micro-data level, and the resulting dataset is treated as if it is a sample from one population. RESULTS: For the separate approach, it is recommended that the simple average of the estimates be used. For the pooled approach, it is recommended that weights be scaled by a constant factor where a period estimate covering the time periods of the individual cycles can be created. The choice of method depends on the aim of the analysis and the availability of data. INTERPRETATION: Combining cycles should be considered only if the most current period estimates do not suffice. Both methods will obscure cycle-to-cycle trends and will not reveal changing behaviours related to public health initiatives.
BACKGROUND: A single cycle of the Canadian Community Health Survey (CCHS) may not meet researchers' analytical needs. This article presents methods of combining CCHS cycles and discusses issues to consider if these data are to be combined. An empirical example illustrates the proposed methods. DATA AND METHODS: Two methods can be used to combine CCHS cycles: the separate approach and the pooled approach. With the separate approach, estimates are calculated for each cycle separately and then combined. The pooled approach combines data at the micro-data level, and the resulting dataset is treated as if it is a sample from one population. RESULTS: For the separate approach, it is recommended that the simple average of the estimates be used. For the pooled approach, it is recommended that weights be scaled by a constant factor where a period estimate covering the time periods of the individual cycles can be created. The choice of method depends on the aim of the analysis and the availability of data. INTERPRETATION: Combining cycles should be considered only if the most current period estimates do not suffice. Both methods will obscure cycle-to-cycle trends and will not reveal changing behaviours related to public health initiatives.
Authors: Maegan V Mazereeuw; Alexander Yurkiewich; Sehar Jamal; Caroline Cawley; Carmen R Jones; Loraine D Marrett Journal: Health Promot Chronic Dis Prev Can Date: 2017-06 Impact factor: 3.240
Authors: Susan Quach; Jemila S Hamid; Jennifer A Pereira; Christine L Heidebrecht; Shelley L Deeks; Natasha S Crowcroft; Sherman D Quan; Stephanie Brien; Jeffrey C Kwong Journal: CMAJ Date: 2012-09-10 Impact factor: 8.262
Authors: Sarah A Buchan; Laura C Rosella; Michael Finkelstein; David Juurlink; Jennifer Isenor; Fawziah Marra; Anik Patel; Margaret L Russell; Susan Quach; Nancy Waite; Jeffrey C Kwong Journal: CMAJ Date: 2016-08-08 Impact factor: 8.262
Authors: Scott B Patten; Jeanne V A Williams; Dina H Lavorato; Jian Li Wang; Andrew G M Bulloch; Tolulope Sajobi Journal: Soc Psychiatry Psychiatr Epidemiol Date: 2016-01-07 Impact factor: 4.328