Using Proxy Measures and Other Correlates of Survey Outcomes to Adjust for Non-Response: Examples from Multiple Surveys