Big Five was created through statistical analysis - they took a whole bunch of traits / questions and figured out that certain ones correlated with other ones.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4104167/
Of course the problem is that Big Five has no theoretical basis, unlike socionics. But it does show that certain traits are essentially bimodal - not that that has any bearing on socionics, which measures different traits that haven't been empirically investigated with any depth.