Structure Harvester: Highest Delta K, not highest probability.

[Miranda Sherlock]

Hello,

I am doing some pop gen analysis on 4 populations of cichlid fishes from 4 lakes. My structure harvester results indicate that K=2 is best for the data (delta K is by far the highest), yet the K value with the highest probability is 5. The variance of K values seems to increase with K, so is delta K a reliable way to select a K value in this case? Sorry for what is probably a very basic question!

Both are plausible values for the population, given that one population seems to be significantly distinct (as seen from prelim. mtDNA analysis and other pop gen methods), however K=5 seems to portray this population as more 'distinct'.

Thanks,

Miranda

[Andrea Schreier]

Hi, Miranda! In my experience, Delta K often misses fine scale structure. Is it possible that you have hierarchical structure among your lakes? For example, perhaps Delta K is capturing the genetic divergence between that significantly distinct population and everything else. There may be substructure among the four other populations which is what is reflected in your plot of L(K). Does that seem biologically plausible?

Andrea  

--
You received this message because you are subscribed to the Google Groups "structure-software" group.
To unsubscribe from this group and stop receiving emails from it, send an email to structure-software+unsub...@googlegroups.com.
To post to this group, send email to structure-software@googlegroups.com.
Visit this group at https://groups.google.com/group/structure-software.
For more options, visit https://groups.google.com/d/optout.

[Berthold Heinze]

Dear Miranda,

you seem to have very high variability (std. Deviation) at e.g. K 3 - is it possible that there are some runs that do not converge and thus raise variability? This could influence the delta K result.

Best regards,

Berthold

--

[Miranda Sherlock]

Hello,

This is definitely possible. I thought that with 500,000 MCMCs and a burn in of 50,000 this would be unlikely? This is my Msci project so I am quite new to data analysis haha!

On Monday, 20 February 2017 17:18:57 UTC, Berthold wrote:

Dear Miranda,

you seem to have very high variability (std. Deviation) at e.g. K 3 - is it possible that there are some runs that do not converge and thus raise variability? This could influence the delta K result.

Best regards,

Berthold

Am 20.02.2017 12:19 nachm. schrieb "Miranda Sherlock" <miranda...@gmail.com>:

Hello,

I am doing some pop gen analysis on 4 populations of cichlid fishes from 4 lakes. My structure harvester results indicate that K=2 is best for the data (delta K is by far the highest), yet the K value with the highest probability is 5. The variance of K values seems to increase with K, so is delta K a reliable way to select a K value in this case? Sorry for what is probably a very basic question!

Both are plausible values for the population, given that one population seems to be significantly distinct (as seen from prelim. mtDNA analysis and other pop gen methods), however K=5 seems to portray this population as more 'distinct'.

Thanks,

Miranda

--
You received this message because you are subscribed to the Google Groups "structure-software" group.
To unsubscribe from this group and stop receiving emails from it, send an email to structure-software+unsub...@googlegroups.com.

To post to this group, send email to structure...@googlegroups.com.

[Miranda Sherlock]

Hello,

That sounds plausible for the biology and something I hadn't even thought of! 

Thanks,

Miranda

On Monday, 20 February 2017 16:41:12 UTC, Andrea wrote:

Hi, Miranda! In my experience, Delta K often misses fine scale structure. Is it possible that you have hierarchical structure among your lakes? For example, perhaps Delta K is capturing the genetic divergence between that significantly distinct population and everything else. There may be substructure among the four other populations which is what is reflected in your plot of L(K). Does that seem biologically plausible?

Andrea  

On Mon, Feb 20, 2017 at 3:19 AM, Miranda Sherlock <miranda...@gmail.com> wrote:

Hello,

I am doing some pop gen analysis on 4 populations of cichlid fishes from 4 lakes. My structure harvester results indicate that K=2 is best for the data (delta K is by far the highest), yet the K value with the highest probability is 5. The variance of K values seems to increase with K, so is delta K a reliable way to select a K value in this case? Sorry for what is probably a very basic question!

Both are plausible values for the population, given that one population seems to be significantly distinct (as seen from prelim. mtDNA analysis and other pop gen methods), however K=5 seems to portray this population as more 'distinct'.

Thanks,

Miranda

--
You received this message because you are subscribed to the Google Groups "structure-software" group.
To unsubscribe from this group and stop receiving emails from it, send an email to structure-software+unsub...@googlegroups.com.

To post to this group, send email to structure...@googlegroups.com.