The quality and the adequacy of data can be a real threat to the kind of research that we want to do. Typically, social services research is going for hard-to-reach populations, so sometimes we can't access and retain and recruit all the study participants that we want, so we may have some missing data arising from not being able to interview everyone we want. Also, if we 're relying on records, agency records, for sources of data, social services agencies, particularly those in safety net systems of care, tend to have pretty weak infrastructure, so that can contribute, as well, to missing data.

What I have learned about handling missing data is that you need to work with experts. I always team with people who have expertise in data management. There is real skill in data management. How do you collect, protect the quality of, clean, store, and make sure that your research team doesn't contribute to missing data? So turning to people who have expertise in data management is really critical.

Then there are particularly specialized ways of analyzing and handling missing data. For handling it, there are a growing number of imputational methods. The choice of an imputational method often depends on the nature of the missingness. Why is the data missing? Is it missing at random, or is there some systematic reason that contributes to missing data? And then there are ways of imputing data or using the existing data that you have to make estimates that are reliable and valid to hold the place of that missing data, but this is time-consuming and requires some very specialized techniques, so I work with statisticians and data managers who know the array of ways of imputing and then how to handle that in statistical analysis.