Boostrapping / Confidence Intervals - IM vs No migration model.

[Kevin Hawkins]

Ryan + others,

     I have been working with dadi for a little while now, and I am still somewhat confused as to how to bootstrap and get confidence intervals. Our data is RAD-seq data and we are testing the difference between a no-migration model (where the migration params are set to 0) and a migration model. I ran both models multiple times to allow theta, and LL to converge, I utilized a LL ratio test to determine which model was more consistent with the data. I am not sure if A: this is the proper approach to determine if there is a significant difference in model selection (Migration vs No migration) and B: if there is a way to obtain confidence intervals. 
After reading the manual as well as some of the other replies on the group, I am still somewhat confused as two how to obtain bootstrap datasets as well as since our data is a genomic scan, if the Hessian approach is appropriate. I apologize if this seems like a very broad and trivial question. 

Thanks, in advance, for your help 

[Gutenkunst, Ryan N - (rgutenk)]

Hello Kevin,

Technically, the LRT assumes the SNPs are independent, which probably isn't strictly true for your data (particularly if you took more than one SNP per RAD-seq locus). If the difference in likelihoods is large, I wouldn't worry about and would just report the results with the caveat that you may be underestimating the p-value because of non-independence. If the difference is marginal, then you need to be more careful. You can do an LRT with a parametric bootstrap like we did in the PLoS Genetics paper, but it's a lot of work. The other option is to downsample your SNPs to just 1 per RAD-seq locus, which makes it easier to argue for independence.

Parameter confidence intervals are similar, although in that case a conventional bootstrap suffices, which is much easier to set up than the parameter bootstrap needed for the LRT.

We've recently developed some Godambe Information approaches that will deal with the linkage without bootstrapping. If you'd like to go that way, I can put you in touch with the relevant member of my group (Alec Coffman, CC'd).

Best,

Ryan

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

--

Ryan Gutenkunst

Assistant Professor
Molecular and Cellular Biology
University of Arizona
phone: (520) 626-0569, office LSS 325

http://gutengroup.mcb.arizona.edu