Yeah but part of good scientific practice is to isolate the variables as much as possible with the control groups. One can work with less controlled variables but it increases noise generally and should be avoided as it is more likely to produce entanglements that fool even the most honest of scientists. The data tends to be noisy and messy when it involves free human populations in even harder sciences with tight qualitatives.

Before blood-typing trying to go with an ethnic system is fair enough as a guess and could potentially be a slightly better match rate (more likely to have similar blood types - especially in more homogeneous populations). There are better tools and groupings now so it raises the question 'why bother with a rough proxy when there are more precise alternatives'?

If looking at just light level one doesn't just go with 'night' or 'day' for light levels when specific times for measurements are possible, controlled temperature and calibrated light controlled rooms would specifically testing for a /different/ aspect and even then the point is to try to figure out what is the missing aspect from the more controlled simulation.

Sickle cell anemia may be found more in those of more immediate African descent but taking prevalence of oxygen carried by whole population and concluding African blood carries less oxygen because some have sickle cell anemia is unscientific. One should be grouping by what is actually germane - characteristics of the blood itself and how it is capable of holding more or less oxygen. Not the pigmentation of the sources.