Admixtools 2

[Kale]

Admixtools 2.0.8 Includes fitted F4-statistics / residuals / gendstats (whichever term you are familiar with) for qpadm!

[Genetics189291]

Admixtools1 vs 2

I guess you can't really take admixtools 2 seriously. It's too much of a difference between the two, I'm using the same outgroups for both

[AimSmall]

What versions of each are your running and what environment?

[Gats]

Admixtools1 vs 2

I guess you can't really take admixtools 2 seriously. It's too much of a difference between the two, I'm using the same outgroups for both

Except your right and left pops are garbage... Dzudzuana is low coverage in new and that right pops just look awful imagine having Denisova there, and sure UstIshim is the first outgroup instead of African like Mbuti...

[Genetics189291]

What versions of each are your running and what environment?

Admixtools1 on opensuse Linux latest version and I updated admixtools2 to the latest version on windows . To be fair this could explain why my son is scoring extra east med in admixtools2 because it’s not entirely reliable I’ll run his results through admixtools1

[Genetics189291]

Admixtools1 vs 2

I guess you can't really take admixtools 2 seriously. It's too much of a difference between the two, I'm using the same outgroups for both

Except your right and left pops are garbage... Dzudzuana is low coverage in new and that right pops just look awful imagine having Denisova there, and sure UstIshim is the first outgroup instead of African like Mbuti...

We are talking about the difference in rights and lefts but the results, regardless of whether you think they’re bad or not the output should be near identical to each other. Stop with the passive aggressiveness and keep to the topic

Admixtools1 doesn’t work with mbuti as the first outgroup, it requires it as the first outgroup

[Gats]

Admixtools1 depends on first outgroup while Admixtools2 usually doesn't, not having Mbuti there crashes the model

[Genetics189291]

Admixtools1 vs 2

I guess you can't really take admixtools 2 seriously. It's too much of a difference between the two, I'm using the same outgroups for both

Except your right and left pops are garbage... Dzudzuana is low coverage in new and that right pops just look awful imagine having Denisova there, and sure UstIshim is the first outgroup instead of African like Mbuti...

You sure do seem confident in yourself

[Genetics189291]

Admixtools1 depends on first outgroup while Admixtools2 usually doesn't, not having Mbuti there crashes the model

You should know that with the way you are speaking

[Gats]

Admixtools1 vs 2

I guess you can't really take admixtools 2 seriously. It's too much of a difference between the two, I'm using the same outgroups for both

Except your right and left pops are garbage... Dzudzuana is low coverage in new and that right pops just look awful imagine having Denisova there, and sure UstIshim is the first outgroup instead of African like Mbuti...

You sure do seem confident in yourself

It depends on the left pops if Denisova/Neanderthal should be in right or not. Your model clearly doesn't require any. And normally when people run you never see Denisova there unless modelling Neanderthal ancestry

[crashdoc]

It is easy to have different behavior in admixtools1 vs admixtools2, especially with low coverage samples.

To get similar results, indeed you have to use the same right pops in both versions, but additionally, ensure that low coverage is treated the same way.

As a general comparison you can look here: https://uqrmaie1.github.io/admixtools/ar...mixtools-2

The most important things to take into considerations for low coverage are:

-What's your first parameter in the admixtools2 qpadm command?  Is it a genotype file prefix or precomputed f2data?
-If it is a genotype file prefix, you need to use allsnps = TRUE
-If it is precomputed f2 blocks, when computing those f2 blocks in a previous step, you need to use maxmiss = 1 and afprod = TRUE

[Genetics189291]

It is easy to have different behavior in admixtools1 vs admixtools2, especially with low coverage samples.

To get similar results, indeed you have to use the same right pops in both versions, but additionally, ensure that low coverage is treated the same way.

As a general comparison you can look here: https://uqrmaie1.github.io/admixtools/ar...mixtools-2

The most important things to take into considerations for low coverage are:

-What's your first parameter in the admixtools2 qpadm command?  Is it a genotype file prefix or precomputed f2data?
-If it is a genotype file prefix, you need to use allsnps = TRUE
-If it is precomputed f2 blocks, when computing those f2 blocks in a previous step, you need to use maxmiss = 1 and afprod = TRUE

I did that here is another set of results for my son same outgroups and left populations. Their is even one model with a greek source that is feasible and passes on admixtools2 which is infeasible in admixtools1 and fails but this passes on both with different results. Admixtools1 looking different for the percentage of the Sardinian component with it contributing 27.7 percent in admixtools 1 and him having less guanche

[crashdoc]

It is easy to have different behavior in admixtools1 vs admixtools2, especially with low coverage samples.

To get similar results, indeed you have to use the same right pops in both versions, but additionally, ensure that low coverage is treated the same way.

As a general comparison you can look here: https://uqrmaie1.github.io/admixtools/ar...mixtools-2

The most important things to take into considerations for low coverage are:

-What's your first parameter in the admixtools2 qpadm command?  Is it a genotype file prefix or precomputed f2data?
-If it is a genotype file prefix, you need to use allsnps = TRUE
-If it is precomputed f2 blocks, when computing those f2 blocks in a previous step, you need to use maxmiss = 1 and afprod = TRUE

I did that here is another set of results for my son same outgroups and left populations. Their is even one model with a greek source that is feasible and passes on admixtools2 which is infeasible in admixtools1 and fails but this passes on both with different results. Admixtools1 looking different for the percentage of the Sardinian component with it contributing 27.7 percent in admixtools 1 and him having less guanche

The results will never be exactly the same as they are computing the statistics differently, but at least they should be near as they seem to be in your example.

One more thing you can try when using the genotype file as the first argument is adding f4mode = FALSE

[Genetics189291]

It is easy to have different behavior in admixtools1 vs admixtools2, especially with low coverage samples.

To get similar results, indeed you have to use the same right pops in both versions, but additionally, ensure that low coverage is treated the same way.

As a general comparison you can look here: https://uqrmaie1.github.io/admixtools/ar...mixtools-2

The most important things to take into considerations for low coverage are:

-What's your first parameter in the admixtools2 qpadm command?  Is it a genotype file prefix or precomputed f2data?
-If it is a genotype file prefix, you need to use allsnps = TRUE
-If it is precomputed f2 blocks, when computing those f2 blocks in a previous step, you need to use maxmiss = 1 and afprod = TRUE

I did that here is another set of results for my son same outgroups and left populations. Their is even one model with a greek source that is feasible and passes on admixtools2 which is infeasible in admixtools1 and fails but this passes on both with different results. Admixtools1 looking different for the percentage of the Sardinian component with it contributing 27.7 percent in admixtools 1 and him having less guanche

The results will never be exactly the same as they are computing the statistics differently, but at least they should be near as they seem to be in your example.

One more thing you can try when using the genotype file as the first argument is adding f4mode = FALSE

I’ll try this out when I get home from work  and see if it makes a difference

[Genetics189291]

It is easy to have different behavior in admixtools1 vs admixtools2, especially with low coverage samples.

To get similar results, indeed you have to use the same right pops in both versions, but additionally, ensure that low coverage is treated the same way.

As a general comparison you can look here: https://uqrmaie1.github.io/admixtools/ar...mixtools-2

The most important things to take into considerations for low coverage are:

-What's your first parameter in the admixtools2 qpadm command?  Is it a genotype file prefix or precomputed f2data?
-If it is a genotype file prefix, you need to use allsnps = TRUE
-If it is precomputed f2 blocks, when computing those f2 blocks in a previous step, you need to use maxmiss = 1 and afprod = TRUE

I did that here is another set of results for my son same outgroups and left populations. Their is even one model with a greek source that is feasible and passes on admixtools2 which is infeasible in admixtools1 and fails but this passes on both with different results. Admixtools1 looking different for the percentage of the Sardinian component with it contributing 27.7 percent in admixtools 1 and him having less guanche

The results will never be exactly the same as they are computing the statistics differently, but at least they should be near as they seem to be in your example.

One more thing you can try when using the genotype file as the first argument is adding f4mode = FALSE

Thanks it was the allsnps I had to set it to false. Results are now matching