Referring to the question in the link below :

How can I use Propensity Scores to adjust for survey non-response bias?

I read the paper published by Lee and Valliant (2009) and the pew research method article.

To make things clear, when dealing with multi-stage survey designs, the best weighting method with the least bias is a combination of the propensity adjustment method and the raking method. However, I see that when referring to the propensity method, we need to have a reference survey and combine it with our survey to be able to apply this method. In many developped countries, such reference surveys are available, but what about when applying surveys within countries where such reference surveys do not exist? what should we do? Do we use only the raking method in that case eventhough we know that it shall produce bigger standard errors resulting in less reliable results when making inferences to the target population? Or is there an option to handle such a problem?

P.S. the definition of the reference survey : a reference survey is a survey that is conducted parallel to the Web survey (meaning the survey we are working on) (Lee 2006). The reference survey is required to have more desirable coverage and sampling properties and higher response rates than the Web survey.




Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.