I calculated pairwise Dest values using SMOGD software (Crawford, 2010), but there are no associated p-values to determine significance. I looked a number of papers where they mostly don't state significance, but some do (but aren't well explained). Are you supposed to state significance with DEST?
I would like to compare my values with pairwise FST.
Thanks... however, I am not very experienced in R. Can you just upload your pairwise Dest values from SMOGD and calculate p-values in R from those? Do you have an example script available?
I don't have scripts. Been a while since i used it. I believe you can upload a genepop file as you do for SMOGD. The manual is straight forward.
The best way is using DEMETICs in R, personally i used both methods (SMOGD and DEMETICs) and i found the results of DEMETICs are excellent since it estimates p-value and confidence intervals using bootstrap resamplings (however it calculates both Dest and Fst for conservative comparaison)
So I am having some issues with DEMEtics - half way through the run, it stops and gives me an error (I forgot to write the error down) and in addition, those comparisons that did run, many resulted in N/A ? What does that mean? Does it have an issue with a lot of missing data?
Not sure if you're still wanting an answer for this one but you can also use genodive (genetics program for macs) to calculate P-values for D-est
Did you find a fix for your p-value problem yet Rachel?
I have the same problem with DEMEtics returning 'NA' as the p-value as for a few loci and the global estimate of D. It seems that the issue is sensitive to the number of bootstraps, as while all loci and the global D.est are numeric p-values up to 85 bootstraps, beyond that a couple of loci start to return 'NA' as a p-value (which turns the global Dest p-value to 'NA' too).
I have very little missing data (microsat genotypes) and most of the loci provide numeric p-values at 1000 bootstraps, so I'm not sure what is causing this failure, and I can't spot any indication in the DEMEtics manual or the Manly book cited by it.
Any assistance appreciated!
Recent versions of GenAlEx can also do this, and might be more user friendly than the ones suggested so far. SMOGD is of course useless when you want to compare genetic structure for different groups, but bear in mind that p-values are quite unpopular among many statisticians (http://www.bmj.com/rapid-response/2011/11/02/p-values-can-be-misleading). Confidence intervals are generally considered more useful, e.g. if a particular CI includes zero, then there is no structure, and if CIs determined for two groups don't overlap, than the one is more structured than the other. If you prefer to compare your groups using confidence intervals, GENODIVE can do this (but the last time I checked, this option wasn't yet available for all the genetic structure statistics it can calculate).
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics