I am using a national telephone survey conducted every year by the CDC called the Behavioral Risk Factor Surveillance System (BRFSS) to answer a question about breast cancer screening rate in low-income counties vs. higher income counties in one state. The number of respondents in one year from the low-income counties is rather low, so I am interested in aggregating the survey data from 2011, 2012, and 2013. The survey methodology did not change substantially in that time.

Here is a CDC document on the BRFSS methodology: http://www.cdc.gov/brfss/data_documentation/PDF/UserguideJune2013.pdf

My question is: are there statistical issues if I aggregate the three years of data to have a larger sample of data? I will need to calculate crude screening proportions but also eventually use logistic regression as well to control for socioeconomic confounders. I appreciate your help!

  • $\begingroup$ for a definitive answer, ask here.. and tell us what you find! $\endgroup$ – Anthony Damico Sep 10 '14 at 15:26
  • $\begingroup$ OK I just did. I'll post their answer when I get it. $\endgroup$ – JJM Sep 10 '14 at 16:43
  • $\begingroup$ Unfortunately their reply was pretty terse, so if anyone would like to answer this question with a more thorough explanation, I'm sure it would be helpful to some people. Here's the CDC's response: "You can use SMART county data to combine multiple years of data to conduct analysis. It’s feasible." $\endgroup$ – JJM Sep 12 '14 at 13:14
  • $\begingroup$ ..how does that reply not answer your question? $\endgroup$ – Anthony Damico Sep 12 '14 at 16:30

You can do this, but you need to account for it when applying the survey weights. My understanding is that if you are only combining years in order to increase sample size (i.e. you are not looking at year effects), then you need to revise the weights, most simply by dividing them by the number of years you are combining. In your case, you would divide by 3. The population you are now generalizing to is the average population over those three years.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.