So, your concern & anxiety is why non-African men are not showing the same improvement of the African men? Are you then assuming all non-Africans are doping, and if so, all at the same dosing?
What about testing in Africa vs testing, in say, the European nations? I haven't seen any Africans popped in OOC testing during the 00s - it's been all IC. OTOH, there's been some EPO positives for European athletes invloving OOC testing. Could the risk of OOC testing within the European nations during the 00s influence some athletes to use smaller, less effective doses or to not dope at all? I'm not convinced that everyone in Europe, America, Oceanian, etc was doping. You've painted with a broad bush the assumption that all the athletes from these continents were doping, hence why can't they match the faster times by the Africans.
IMO, I think a more practical way to measure the effectiveness of rEPO (and essentially that's what trying to do with your statistical analysis in the first place - with the assumption everyone post-90s was doping) is to look at individual positive cases and measure improvement from baseline. Doping positives & sanctions are no-brainers (I include blood doping cases as rEPO & blood transfusions achieve the same objective - raising Hct, i.e. "O2-vector doping"). Athletes caught up in doping scandals & doping rings could also be used (e.g., Operation Galgo). Athletes training under a known coach and/or agent with ties to doping would also be fair (e.g., Aden & Rosa). However, steroid positives with male distance runners would be difficult. Do we lump them in with the O2-vector group under the pretense that EPO/blood doping & roids is a popular combination with endurance athletes, or do we let the roid positive stand on it's own?
Again, IMO, examining individual doping cases would be a more practical way than to make the broad assumption that EVERY athlete post-90s was doping with rEPO/blood tranfusions.