this may seem like a silly question to someone who knows the mathematics of these parameters better than i, but i'm aware Gst and Fst can be presented as a proportion of their theoretical maximum where appropriate, but can Rst?
I ask because i am calculating both for haploid microsatellite data, and i recently calculated their theoretical maximum each from recoded data of the original dataset. for the original data the Fst and Rst (calculated in Arlequin) previously gave me exactly the same results (barring negative values of Fst which were of course 0 in Rst), but when the recoded data values where fed to arlequin (to get the theoretical maximum), Fst gave me values from 0.5 to 0.6 (which seem reasonable), while Rst gave me values in excess of 1.3. could anybody explain why this may have occurred?
Thanks.
Regarding this question: Yes, there’s no reason you can’t per se. However, remember you’re standardizing based on the variances and not based on heterozygosities. They are related to heterozygosities, in that they’re both related to coalescence times, but the expected value of the variance of allele sizes is a product of that (and other things) with mutation rate. The mutation rate just so happens to fall out in the end in RST, but it means you can’t standardize on heterozygosity itself. If you already are, I can’t help you further because I’m not sure what you’re doing in Arlequin to standardize. In any event, I’d be standardizing by hand this this is a very non-standard application.
Lorenzo Zane gave you an excellent answer to your previous question regarding Rst, which I would refer you to again. I’d add to his answer that well studied systems tend to shy away from strict SMM like behavior, and more towards TPM behavior, which makes me (and most other people) leery of RST. Small deviations from SMM can lead to problems. Additionally: Why you feel the need to standardize, and why RST? If you feel that FST is underestimating differentiation due to high diversity, moving to RST is not a clear solution. Either use Hedric’s standardization (F’ST) or Lou Jost’s D, which is, in my opinion, superior in its derivation and easier to interpret to boot (it shows diversity partitioning, full stop). And as the variances in allele size aren't bound by 1 and 0, I'd be hard pressed to imagine a situation where I'd want to standardize on them anyhow.
HI Thomas
Indeed, Kevin gave you an excellent answer. However, I tend to disagree in just one thing. It is true that most well studied systems are best explained by the TPM model. However, the TPM is usually closer to the SMM than to the IAM (i.e., about 75-90% of the event occur according to the SMM). Therefore using the SMM could be better than using the IAM (for example, if you want to calculate migration rates among populations)..
Regarding your question about Arlequin's results, could be the errors of your variance estimates are too large? Anyhow, as Kevin told you, I would also use Jost's D.
hope this helps
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools