Elimination of outliers when calculating population averages on GEDmatch

[reboun]

How accurate is to eliminate the outliers when calculating a population’s autosomal structure? Think of a population which is really diverse and 40% of the population is in the outlier range. Isn’t 40% is a huge percentage to neglect?

[Lemminkäinen]

How accurate is to eliminate the outliers when calculating a population’s autosomal structure? Think of a population which is really diverse and 40% of the population is in the outlier range. Isn’t 40% is a huge percentage to neglect?

It depends on the sampling. The sampling can be biased, so you don't know the reality. If you see extreme diversity you can assume that the problem exists, but not the size.

[reboun]

It depends on the sampling. The sampling can be biased, so you don't know the reality. If you see extreme diversity you can assume that the problem exists, but not the size.

Yes, but elimination of outliers make a diverse population seems less diverse than it is. I guess it is inaccurate. Some populations are very diverse thanks to the fact that they have always been on immigration roots of tribes.

[Jana]

Which population consists of 40% of outliers? Suburbs or Paris and Brussels, you mean?

I know what you are aiming for though and it's simply innacurate. 40% of ethnic Turks certainly aren't genetic outliers. Balkan Turks aren't outliers and have typical for them genetic structure.
You are zero ethnic Turk and thus suffer from identity crisis.

Makes no sense to include people like you in autosomal study of Turkish population. And if you are implying 40% of Turks in total are consisted of foreigners like Bosniaks, that is obviously nonsense.

[reboun]

Which population consists of 40% of outliers? Suburbs or Paris and Brussels, you mean?

You are zero ethnic Turk and thus suffer from identity crisis.


Okay, this is your opinion, which I respect. By the way, it might sound ignorant but what about suburbs of Paris and Brussels?

[Leto]

Can you please finally stop with this bullshit? You don't have to open a thread for every nonsensical question that comes to your mind.

[princeton90]

Contrarily, elimination of outliers increases the accuracy. If a population has 40% of outliers then that 40% is the percentage of ethnic minorities in the population for sure.

[reboun]

And if you are implying 40% of Turks in total are consisted of foreigners like Bosniaks, that is obviously nonsense.

Stearsolina, I don’t think it is nonsense. About 40% of Turkish population might consist of people with non-Turkic and non-native Anatolian background.

[mentalanesthesia]

Stearsolina, I don’t think it is nonsense. About 40% of Turkish population might consist of people with non-Turkic and non-native Anatolian background.

Anatolian Turks, Balkan Turks and other Turkic minorities in Turkey are ethnic Turks. Non-Turkic minorities are not outliers because they are not part of the ethnic Turkish population. They are Turkish citizens not ethnic Turks. That 40% seems inflated. Do you include all Balkan Turks in that number. Also you can identify as a Turk and nobody would reject you as a Turk but you shouldn't overdo the diversity part by including different ethnicities as Turks. They have their own different identities but the only identity for the Anatolian Turks is Turkish.

[reboun]

Anatolian Turks, Balkan Turks and other Turkic minorities in Turkey are ethnic Turks. Non-Turkic minorities are not outliers because they are not part of the ethnic Turkish population. They are Turkish citizens not ethnic Turks. That 40% seems inflated. Do you include all Balkan Turks in that number. Also you can identify as a Turk and nobody would reject you as a Turk but you shouldn't overdo the diversity part by including different ethnicities as Turks. They have their own different identities but the only identity for the Anatolian Turks is Turkish.

I thought of people who have non-Turkic ancestors in last 2-3 generations. I guess they make 40% of Turkish population or no?

[Dunai]

I had the same dilemma when making the Hungarian regional averages, but if someone has largely ethnic Hungarians (85%+) among their great grandparents than even if that person is an outlier compared to other ethnic Hungarians, I did include them, since outliers will exist all the time, and why should that person's result be considered less "Hungarian-like" while most of his or her great grandparents were indeed ethnic Hungarians? I guess same can be applied to all other ethnic groups, including Turkish people. When you exclude outliers you are cherry picking and the results will thus not be as representative than with outliers included.

[mentalanesthesia]

I thought of people who have non-Turkic ancestors in last 2-3 generations. I guess they make 40% of Turkish population or no?

Can you elaborate on that 40 percent? Which peoples constitute the 40 percent non-Turkic "Turkish" population?

[placebo]

I thought of people who have non-Turkic ancestors in last 2-3 generations. I guess they make 40% of Turkish population or no?

Impossible in my opinion. Not more than 30% maybe %20. (thanks to the kurds, they're 11-15 million)

[reboun]

Can you elaborate on that 40 percent? Which peoples constitute the 40 percent non-Turkic "Turkish" population?

I mean people with known non-Turkic speaking background in last 2-3 generations, such as Bosniak, Arabic, Circassian and etc. Maybe not 40% but my observations tell me that people with such roots are common.

[reboun]

Can you elaborate on that 40 percent? Which peoples constitute the 40 percent non-Turkic "Turkish" population?


Impossible in my opinion. Not more than 30% maybe %20. (thanks to the kurds, they're 11-15 million)

Yes, 20-30% sounds more reasonable than 40%.

[mentalanesthesia]

Yes, 20-30% sounds more reasonable than 40%.

Abi Türküm de geç çok kafana takma bu meseleleri. Büyük ihtimalle gelecekte Türkiye'nin nüfusu daha da homojenleşecek büyük şehirlerde devam eden karışmalarla. Ama bu süreç şu an daha tamamlanmadı ve halklar arasındaki sınırlar hala duruyor.

[reboun]

When you exclude outliers you are cherry picking and the results will thus not be as representative than with outliers included.

I completely agree. Excluding outliers is nothing but a kind of cherry-picking.

[RogueState]

Sample manipulation, unweighted samples, etc. are all traditional techniques of data manipulation (not only in the case of genetics, but everywhere)

If you are interested technically on how to remove outliers, take a statistical course or book on Robust statistics

[princeton90]

I had the same dilemma when making the Hungarian regional averages, but if someone has largely ethnic Hungarians (85%+) among their great grandparents than even if that person is an outlier compared to other ethnic Hungarians, I did include them, since outliers will exist all the time, and why should that person's result be considered less "Hungarian-like" while most of his or her great grandparents were indeed ethnic Hungarians? I guess same can be applied to all other ethnic groups, including Turkish people. When you exclude outliers you are cherry picking and the results will thus not be as representative than with outliers included.

Elimination of outliers is not cherry-picking, it's done in almost all statistical calculations. If one's DNA is an outlier then he or she clearly has foreign ancestry.

[reboun]

Elimination of outliers is not cherry-picking, it's done in almost all statistical calculations. If one's DNA is an outlier then he or she clearly has foreign ancestry.

Elimination of outliers is cherry-picking, period.

[reboun]

Edited