Ever since I tested with 23andMe back in 2015, I have had a small percentage of East African admixture in my Ancestry Composition report. My East African result looked like this:
The Timeline screenshot at the bottom shows that although the total percentage was very small and the segment on chromosome 6 very short, 23andMe estimated my East African ancestor to have been born, most likely, between the years 1720 and 1810 (but possibly earlier). In a genealogical perspective, this is recent and well within the timeframe covered by civil and church records in most Western countries.
In April this year, however, 23andMe discontinued the <0.1% category, rounding it up to 0.1% or down to 0.0%. I had <0.1% East Asian, which must have been at least 0.05% because it got rounded up to 0.1%. It is still present in my Ancestry Composition report. However, my <0.1% East African must have been less than 0.05% and was therefore rounded down to zero. It has now disappeared from my Ancestry Composition report, and the segment is read as No Data Available (which I find odd in itself, since the appropriate label would be Unassigned).
Does this mean that the segment was never real? I was informed that 23andMe stopped showing <0.1% categories precisely because such small percentages could be statistical noise, and they wanted to err on the side of caution. However, the fact that a segment is small does not in itself mean that it is noise.
My East African segment at 23andMe showed at 50% and 60% confidence, but was read as Unassigned at 70% confidence and higher. In other words, there was always a real possibility that the segment was statistical noise, but the odds were slightly in favour of it being real.
To find out whether or not an admixture segment is real, one can do two things: (1) compare one's own results using different admixture analyses (preferrably visualized through chromosome paintings), and (2) compare one's own results with relatives who match you on the segment in question, and see if those relatives have the same admixture as you in that specific location (I used this method very fruitfully to verify the West African segment on my paternal grandmother's side).
A detailed comparison between GEDmatch chromosome paintings shows that my African segment on chromosome 6 is picked up consistently by different datasets and calculators, which is strong evidence for the segment being real and not statistical noise. The segment's realness is further confirmed by the fact that my mother has African DNA in the exact same location as me, and more than me, which shows that the segment comes from her (my father, as shown in the images below, has no African in this location). The African segment is often read as primarily East African, which confirms 23andMe's assessment.
I have been seriously reflecting on whether I really do have East African DNA (and relatively recent East African ancestry), as 23andMe kept telling me prior to April this year. After reviewing the GEDmatch chromosome paintings, I conclude that the East African segment is real, and that 23andMe were too cautious in this particular case - throwing out the baby with the bath-water, as it were. Since my mother's African segment seems to be a bit longer than mine, it is quite possible that hers might show up at 23andMe if she were to test with them.
I still don't know which side of my mother's family the East African DNA comes from. Finding this out is an ongoing task. It could be connected with our Romani side. The Romani people are known for travelling long distances, and someone of fully or partly East African ancestry might have married into one of our Romani family lines in the early 1700s, for instance; a period where there are a lot of gaps in our Romani ancestral tree. It is possible that the East African DNA has something to do with the Trans-Atlantic slave trade, especially since genealogist and blogger TL Dixon has suggested that my mother's African segment (which is sometimes shown mixed with Asian/Oceanian) might in fact come from Malagasy ancestors. The Romani hypothesis and the Malagasy hypothesis might both be true; or only one of them might be true; or perhaps neither is true. Interestingly, the reference populations in 23andMe's "East Africa" category seem to point in a third direction: The Horn of Africa (Ethiopia and its neighbours). Further research will hopefully solve this puzzle one day, and tell us the identity of our East African ancestors.
Update, 2 June 2018:
I just noticed that 23andMe seems to have made a little "slip" admitting that I do, in fact, have East African admixture:
I get the same wording for my Native American ancestry, which, like my East African, also used to be present at <0.1%.