advanced analytics for survey with low response rate?

[Liz Smith]

HI all, I am working with a survey where the response rate was 10%, so very low. So far I have stuck to descriptive statistics to analyze the data, but I am considering doing a multinomial logit model to see whether certain variables are associated with a dependent variable. My question: does it make sense to do advanced analytics with such a low response rate?

Thank you!

[Dick Campbell]

A low response rate does not necessarily invalidate a survey. Suppose that you could show that the ten percent who responded are in fact a random sample (i.e. a subsample) of the original sample? Would you throw away the data because the response rate is "too low.?" The real question here is bias. I grant that many editors and readers assume a ten percent response rate is clear evidence of bias but I would disagree.So, do you have a reasonable sample of the population you started with? If not, then descriptive statistics suffer from the same problems that would affect a more sophisticated analysis.

[Liz Smith]

This is very helpful, thank you.