Negating a middle category yields a comparison group too heterogeneous to interpret

Evidence

The default is just to do the negation of the gray zone, which means including both positive and negative pole cases in the comparison group. It is not obvious that this is always the right or best choice. Essentially one has the problem of heterogeneity of the comparison group. Imagine a standard statistical analysis where the gray zone variable is X1. To give this content let’s use the common hypothesis that competitive–authoritarian regimes are more prone to have civil wars. The question is, more prone than what? As understood in the literature it is both more war prone than democracies and more war prone than autocracies. — Gary Goertz, Social Science Concepts and Measurement, 1164

Categorization

• XREF: Speaks directly to earlier highlights on comparing democracy to mixed regimes and the ambiguity of the negation case.
• QUESTION: Is there a principled way to choose which pole drives an effect when the control group is split between both extremes?