Tuesday, May 3, 2016

CCHS - 1 and 2 year files


We’re wondering if there is any way to merge / combine the annual release of the Canadian Community Health Survey with the two-year merged file.

Is there any way to merge the 2012 annual CCHS PUMF file with the 2011-12 two year PUMF? I am working with a student who is interested in the variable named CHP_03 on both files. There are many more valid cases on the 2012 file than on the 2011-12 file, however, for other variables she needs it is the reverse, so if it were possible to merge that would greatly help her analysis.

More generally, why does the two-year file, while having twice as many cases overall as the one-year file, sometimes have far fewer valid responses on certain variables? For the above variable, the one-year file (annual component) has 61607 cases while the two year file has only 6943, or about a tenth as many. This seems odd.


CHP was optional content for 2011 and only MB chose it, while in 2012 it was Core content. The optional selection of the module in 2011 will mean more valid cases on the 2012 file because the two year file only has data for those provinces that selected the module in both years.

The 2011-2012 file is already the combination of the two individual years, so trying to append on the 2012 cases again would duplicate those respondents. It would not give you any more cases to analyse. If there is some correlate variable that you want to look at against CHP_03, it should be on the 2012 file. Anything on the two-year file is necessarily on the one year files. There will be more records on the two-year, but they will be the respondents from 2011, which as you know only had data for CHP_03 in Manitoba.

Any pooling of years should be done where the content is common across the same geographies over those years. Maybe consider looking at the 2013-2014 PUMF. CHP was still core for the latest release, so maybe it would support the analysis (and give more recent results).