consistent results between High density SNPs and thinned SNPs

[biozzq]

Hi @bumblenick

I want to know whether the thinning process is necessary before running admixtools when using high density SNPs. I just care about the accurate of the results.
I have run F3, F4 and D-statistics using about 40M SNPs without pruning and 13M SNPs with pruning by LD. When running with 40M SNPs, it also runs fast, and the result is almost consistent with the thinned SNPs. So I want to know, If I just want to run above statistics for a small number of population, can I get a more reasonable result when using all SNPs?

Best
Zhuqing

[bumblenick]

On human data 1M SNPs is adequate for nearly all pop gen data. In my experience thinning makes little difference and in all cases the standard errors are meaningful, In your case I would I think thin down to 1M snps and not worry too much. Exception: You have very large number of samples and want to study very rare alleles; then things may be different. Nick

…

On Tue, Jan 24, 2017 at 12:13 AM, biozzq ***@***.***> wrote: Hi @bumblenick <https://github.com/bumblenick> I want to know whether the thinning process is necessary before running admixtools when using high density SNPs. I just care about the accurate of the results. I have run F3, F4 and D-statistics using about 40M SNPs without pruning and 13M SNPs with pruning by LD. When running with 40M SNPs, it also runs fast, and the result is almost consistent with the thinned SNPs. So I want to know, If I just want to run above statistics for a small number of population, can I get a more reasonable result when using all SNPs? Best Zhuqing — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#14>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AQn_h3E1PIbblNN2HiXFI43HuQOvwoVqks5rVYh2gaJpZM4Lr3xe> .