How Can We Combine Data Sets With an Unequal Number of Categories?

• Activity Number: 320
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: ENAR
• Author(s): Stef van Buuren*+
• Companies: Netherlands Organization for Applied Scientific Research
• Keywords: multiple imputation ; data combination ; measurement ; MICE

Abstract

A common problem in combining datasets is that the numbers of categories of an item may differ across sources. Practitioners typically resort to ad-hoc recoding to the fewest number of categories. However, such ad-hoc recoding actually makes important implicit assumptions about the relations between the different variants. If these assumptions do not hold, the consequences of the seemingly trivial recoding operation on the estimate of interest can be dramatic.

If a double-coded sample is available that contains both category codings, multiple imputation can be used to "translate" the information into another category system, while accounting for the inherent uncertainty of the transformation. I will present a small example that illustrates the key principles and assumptions.