'pair.ia' NA's

[Drew Hardy]

Hello, I am working on using 'poppr' calculate the index of association for some populations of killifish within a hybrid zone and I'm running into an issue that I'm hoping you could help out with.

I'm running the 'pair.ia' command within the 'poppr' package on 2 of our 'reference' populations - 1 for each parental species within the hybrid zone. These reference populations contain putatively 'pure' individuals (genetically) of either parental species with an n = ~30 in each, and thus have very little to no SNP differences between individuals in the same reference population at 200 subsampled loci. For all other populations in the hybrid zone we have sampled (containing both hybrids and parental individuals), we are able to get good data for calculating IA/rD using 'poppr', but for these two parental populations, we are getting a lot of NA's. Specifically for one reference pop, 'G', we got all NA's for every pairwise comparison among loci. For our other reference population (H), we have very high numbers for IA/rD, and the vast majority of pairwise comparisons have NA's associated with them. My question for you is, do these NA's arise simply due to there being absolutely no difference between the genotypes of the 200 loci dataset between all individuals in the population (what we would expect out of an entirely genetically 'pure' parental reference population)? Would we be able to make an assessment as to what value these populations should hold for IA/rD (using 'poppr::ia') for no difference in SNPs? 1.0 for 100% association? 

Thank you in advance for the insight.

Best,

Drew Hardy

University of North Florida

MS Program, Biology

[Zhian Kamvar]

There's a lot going on here so I'm going to answer this the best I can, but I make no promises that it will actually answer your question.

First, the last chapter of my dissertation handles factors that could influence values of the index of association. You can read the dissertation version or you can read the (pdf) version that we intended to publish, but never did (largely due to my moving out of the field), which demonstrates why you can achieve severely negative values of the index of association with an ostensibly clonal population. 

Second, you will receive negative values of the index of association looking at a single pair of loci when one of those loci is monomorphic since the observed and expected variance of that locus will be zero, leading the value of the denominator of rd to be zero and thus leading rd to be not a number. With Ia, the denominator is the sum of the expected variance for all loci, so that value will simply be zero, which makes sense because a locus with no variance can not tell you any information about linkage with another locus. 

Hope that helps.

Best,

Zhian

--
You received this message because you are subscribed to the Google Groups "poppr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to poppr+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/poppr/94d42f1b-64e8-464e-ba9f-3ef168de47a0n%40googlegroups.com.