interpretation of delta k results

[Meghana]

Dear All,

The delta K  for my structure result is highest for K=2, but only by a few decimal points, above K=4. Is it ok to take K=2 to be the final result?

Also, since K=1 will never be selected, how can we be sure that K=1 is not the result? Is there any other parameter (such as L) that we can use instead of delta K?

Thanks in advance.

regards,

Meghana

[Ramesh krishnan Ramasamy]

Dear Meghana,

Hopefully, there is no need to rely only on delta K value. In my experience, most of the studies used LnP(D) or Delta K value. Even, some people used other parameters like results of cluster analysis, PCoA analysis, etc...

Good luck

Ramesh

--
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-softw...@googlegroups.com.
To post to this group, send email to structure...@googlegroups.com.
Visit this group at http://groups.google.com/group/structure-software.
For more options, visit https://groups.google.com/groups/opt_out.

--
Ramesh krishnan R
Junior Research Fellow
Molecular Biology Lab-1
Central Sericultural Research & Training Institute
Ministry of  Textiles, Govt. of India
Srirampura, Manadavadi road
Mysore - 570 008

Mobile: +91 8951696997, +91 7502533038
Email: rame...@gmail.com

[Meghana Natesh]

Dear Ramesh,

thank you for responding. 

I am just wondering which to rely on. The mean LnProb of data has the most negative value for K=1 and the least variance. Delta K suggests K=2 (although K=4 is pretty close). So by Ln P, should I go by K=1?

Biologically I suspect there to be structure. 

I have attached the files, please let me know if you can suggest something.

Thank you

[Ramesh krishnan Ramasamy]

Dear Meghala,

   Its really tough to take right decision in your case. I suggest you to rely mainly on delta K value and then you have confusion only between K=2 or 4. Try both and see, how STRUCTURE biologically assigns the genotypes. The best one you can take for final presentation. I feel biological insight is more important than these number games. Don't forget that some articles even presents and discuss range of k values like 2 to 5.

Hope my answer is little helpful for you!

Good luck!!

Ramesh

[Vikram Chhatre]

The lnPD plot isn't conclusive.  What are the run parameters?  How long was the data collected?

V

[Julie Hebert]

Meghala,

I would recommend taking a look at the histograms for K=2 through K=6, and see if there is any actual information added with the additional K. (In other words, do you each additional color have individuals with nearly full membership in the new color, or do you just see splitting?) I'm pretty sure I've posted some illustrations of what I'm talking about in the archives.

Julie

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.
Visit this group at http://groups.google.com/group/structure-software.
For more options, visit https://groups.google.com/groups/opt_out.

--
Ramesh krishnan R
Junior Research Fellow
Molecular Biology Lab-1
Central Sericultural Research & Training Institute
Ministry of  Textiles, Govt. of India
Srirampura, Manadavadi road
Mysore - 570 008

Mobile: +91 8951696997, +91 7502533038
Email: rame...@gmail.com

--
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.
Visit this group at http://groups.google.com/group/structure-software.
For more options, visit https://groups.google.com/groups/opt_out.

--
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.
Visit this group at http://groups.google.com/group/structure-software.
For more options, visit https://groups.google.com/groups/opt_out.

--
Ramesh krishnan R
Junior Research Fellow
Molecular Biology Lab-1
Central Sericultural Research & Training Institute
Ministry of  Textiles, Govt. of India
Srirampura, Manadavadi road
Mysore - 570 008

Mobile: +91 8951696997, +91 7502533038
Email: rame...@gmail.com

--
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.

[Meghana Natesh]

Hi Vikram,

the runs were pretty long. The burnin was 500,000 and the mcmc chains 1000,000. The alpha plot shows convergence. 

[Vikram Chhatre]

Follow up on Julie's suggestion of comparing bar plots K=2 and above.  You may want to post them all here as well.

V

[Meghana Natesh]

Hi Julie, 

I suppose what you are suggesting is that for any K, the q value should be assymetric, as in each individual should more or less belong to one population overall. That is true for K=2, most individuals are of predominantly a single colour, but I haven't yet checked for other values of K. I will do that now. 

To unsubscribe from this group and stop receiving emails from it, send an email to structure-softw...@googlegroups.com.

[Meghana Natesh]

Thank you for responding!

I will post them soon.