The non-exclusive sets may overlap with interesting results, if there is a multiplicative effect. That's a hypothesis worth testing, perhaps through a quick str

The non-exclusive sets may overlap with interesting results, if there is a multiplicative effect. That's a hypothesis worth testing, perhaps through a quick stratification of the data. There is also an issue about multi-colinearity. For example, if one were to consider modifiers vs nouns. Is it the specific noun that is of interest? Or the subset of that specific nouns, delineated by the modifier? what about the modifier generally, applied to a general noun? etc. Are there specific combinations that are significantly different from the average? etc.