Lost in Translation: The Categorization of Multidimensional Constructs in the Human Sciences

Oregon Convention Center 

There are many important, informative qualities about human subjects that are inherently immeasurable. Even qualities that are measurable, such as age, are often transformed into categories for various statistical design or analysis purposes. Despite scientific evidence to the contrary, biological sex and psychosocial gender are often categorized as binary qualities and only five to eight levels represent different race and ethnicity groups in the US Census. Translating complex, multidimensional constructs into discrete statistical variables inevitably results in information loss. This presentation delves into the statistical issues associated with this kind of information loss. We emphasize human sex as an example where the thoughtless adaptation of traditional categorizations and variable definitions can lead to dire consequences, statistical and otherwise. We also present examples where misinformation due to inappropriate categorization results in misleading statistical conclusions and suggest preventative measures beginning at the data collection stage. Finally, we reflect on the potential impact this has on gender and sex minorities in policy and the health sciences.