How to read Omnipop

[burto]

Hi,
Just got this back from the person in the know who created it!

The E# is the way that Excel displays scientific notation. For example 1,000,000 would be displayed as 1.00E+6 and 1,000,000,000 would be displayed as 1.00E+9. The lower the number, the more common the alleles you entered are found in the particular population.

So now you know!

[Maria_W]

Alot of zeros!

Thank you. Thats very interesting. I can't use Omnipop but would it be the same as some of the other databases like the RCMP and the European database. So 1.00E12 would be in 1.000,000,000,000. How can you get back 1.00E28. It would be insane amount of zeros. Maria

[burto]

Just to add to that you go the LOWEST number not the highest...so 2.16E0 would be better than 9.99E0.
So for me my closest match for all alleles together is.....Argentinian!!! I am a Heinz 57!!!!

[Sonia]

Thanks!

Thanks for posting this information, Burto. One of the things I like about Omnipop is that you can put in one line of alleles and then the other. It still doesn't help me clarify much out of Gram's unusual DNA Tribes results, but I'm starting to see some clustering showing greater odds (if you can call them greater when your talking about 1 in such huge numbers) among certain populations. Based on Gram's descriptions of her parents, I'm almost convinced I know which line of alleles came from each parent. Her dad may have been very dark skinned, but I'm still seeing some other mixing in there.

[Maria_W]

Opposite?

Sonia.
Yes I saw your grandmothers results. Several people said the same thing about my tribes report. Evidently mine are kinda strange and don't fit. Hey, I unique! i seem to be all over the place. I just think its so amazing. I can't use Omnipop. I tryed The Canadian database and the European database. My understanding is that 9.34E7 is better than 2.22E12 for those databases. Sounds like the opposite for Omnipop. 3.34E0 would be better than 4.34E0? Maria

[tomcat]

Omnipop apparently expresses results as CSFS, the Canadians do, as odds. The low E# thingy is universally true. But on ENFSI you want the biggest decimal numbers that come-up.

Tom

[burto]

Have just asked about the reliabilty of statistically based racial prediction and it is reliable 80% of the time. Have also asked about using other databases and comparing to see if that increases the validity if there is a pattern throughout them all. Will let you know the answer.
Also can anyone work out these zero scores...what does E0 mean? For example if you had 2.89E0...what is that...1 in 2.89?

[tomcat]

Yes. E0 is zero exponent, the number is the actual odds of matching another in the population referenced.

[burto]

Just been advised that these databases are not suitable for the kind of matching we are doing. Any ideas?

[tomcat]

On what basis are they not suitable? Or is our use of them unsuitable?

The databases are the same as those employed by DNATribes. Tribes may have a more complete representation than Omnipop, CSFS or ENFSI but Tribes (to the best of my knowledge) is not conducting any field research, they are imply re-purposing data collected by others.

Tom

[Maria_W]

All for nothing

All that work that everyone did trying to figure out thier odds, seems to have been for nothing?.. How do we do it? Maria

[Noaide]

On what basis are they not suitable? Or is our use of them unsuitable?

The databases are the same as those employed by DNATribes. Tribes may have a more complete representation than Omnipop, CSFS or ENFSI but Tribes (to the best of my knowledge) is not conducting any field research, they are imply re-purposing data collected by others.

Tom

The individual matching of your profile against the different populations is the same as DNATribes, that explains why the ranking looks quite the same between Omnipop and DNATribes, you also see this for the ENFSI DB, what make DNATribes different is the continent match analysis part where they match your profile against larger clusters of populations they have identified using the different populations, probably using distribution math or something.

Noaide

[tomcat]

CORRECTION

... Tribes (to the best of my knowledge) is not conducting any field research, they are SIMPLY re-purposing data collected by others.

Noaide,

I understand. And the Tribes Continental/World Region analysis has become much more interesting and useful.

Tom

[burto]

To quote:
"The tests that are used for forensic human identification are not really very appropriate for anything other than a very basic indication of possible major ethnicity, such as Caucasian vs. African (which you probably already know). It is not even always accurate with this type of broad conclusion.
They are definitely not appropriate for the detailed type of genetic genealogy that you are trying to perform"
Said in response to me asking about whether using the different databases to find a correlation would work in finding out the heritage of my American side.
Of course he could just have misunderstood my question or I didn't explain it very well. However he did suggest DNAPrint instead. He has been very helpful.
Maybe someone else could enquire too who can explain a little better how valid these databases are for our purpose?

[Maria_W]

I got Omnipop.

My husband was able to download the Omnipop program for me. Can't wait to see what populations I match too. So it the lowest numbers that I want to match too. O.K. i know that they are not suppose to be suitable but I am still going to do it anyhow. Will report back later. Maria