Merging BED with v62 dataset with Google Colab

[Inquirer]

@TanTin @AimSmall @crashdoc

Does anyone know why Mota is being modelled as partly Eurasian by find_graphs()? I store f2, f3, and f4 statistics in the directory that I feed to the function, by the way. Is that too much?

The graph is otherwise perfect, in my opinion, because it perfectly reflects the splits that I believe occurred between ancient populations (which I added to it in post).

Yes, Mota is  partly Eurasian, because the  Eurasians or OOA came from there, somewhere East Africa. 
BTW your graphs are far from perfect. 
If you think it's perfect,  you should explain why it is so good.  If you ask us questions, you have some or many doubts. 
The connection to East Africa side is generally clear. It is the West Africa / Ghost/ still in the dark.

Yes, but if Mota doesn't have much post-OOA ancestry, then he shouldn't be modelled as ~30% Eurasian. I was aware that he may be ~5% Basal Eurasian, but ~30% is more than I expected.

Also, why is my graph far from perfect?

[TanTin]

Yes, but if Mota doesn't have much post-OOA ancestry, then he shouldn't be modelled as ~30% Eurasian. I was aware that he may be ~5% Basal Eurasian, but ~30% is more than I expected.

Also, why is my graph far from perfect?

The perfect graph will give you the story of all the humans. The perfect graph can explain the origin of AMH and connection to other hominids like Denisovans and Neanderthals. Do you think you are there now ?
Once you have that picture you may other branches like East/ West Africa , OOA, papuans etc..
Do you think you have that one ?

[Inquirer]

Yes, but if Mota doesn't have much post-OOA ancestry, then he shouldn't be modelled as ~30% Eurasian. I was aware that he may be ~5% Basal Eurasian, but ~30% is more than I expected.

Also, why is my graph far from perfect?

The perfect graph will give you the story of all the humans. The perfect graph can explain the origin of AMH and connection to other hominids like Denisovans and Neanderthals.  Do you think you are there now ?
Once you have that picture you may other branches like East/ West Africa , OOA, papuans etc..
Do you think you have that one ?

Well, I meant perfect for my particular goal, which is to produce a model that reflects my hypothetical ancient population splits. I think that the graph achieves that minus its inflation of Mota's Eurasian ancestry (and perhaps minus its elimination of TKH's Eurasian ancestry).

[Inquirer]

@crashdoc, @AimSmall, @TanTin

Are these graphs considered successful models? How can I tell whether or not a graph is successful?

   

   

[Kale]

@crashdoc, @AimSmall, @TanTin

Are these graphs considered successful models? How can I tell whether or not a graph is successful?

There's two criteria to measure whether a graph is sucessful.
1) The f3 residuals ex.
    pop1  pop2                    pop3                        est        se    fit        diff        z      p
    <chr> <chr>                    <chr>                    <dbl>    <dbl>  <dbl>        <dbl>    <dbl>  <dbl>
  1 Mbuti ZlatyKun                ZlatyKun                0.3109 0.0005686 0.3105  0.0003570    0.6278  0.5301
  2 Mbuti ZlatyKun                BachoKiro_IUP            0.1721 0.0006559 0.1715  0.0005773    0.8802  0.3788 
There's no official pass/fail threshold, but I try to keep the Z under |2|

1) Realistic branch lengths, i.e. no implausible 0 edges or implausibly huge edges

Also, as a structural thing, an admixed node should have a drift edge immediately following it in all cases, especially before admixing into something else. It can distort things upstream otherwise.