Thank you for all the answers. There is a listing of all the answers I
received. Unfortunately, it seems that no software is able to compute
automatically C.I. of pairwise FST. I checked all the softwares that
are mentioned in those answers and I can confirm that no one fulfilling
my requirements. I decided to use the solution (the hard and long
one...) proposed in the first answer included in this message. With
a good work technique it takes me about 4 hours (instead of 1-2 day)
to calculated CI for 277 pairwise comparisons. Good luck and many
thanks again!
Philippe.
Original message:
Does somebody know a software allowing the estimation of confidence
intervals for pair-wise Fst using bootstraps over loci? It seems that both
Arlequin 3.01 and Fstat are only able to estimate confidence intervals
for the overall Fst.
Thank you.
===I asked the same question about 8 months ago and received a lot of
software suggestions, including GENEPOP, MSA, SPAGEDI, POPDIST, GENETIX
and the ones you've already mentioned. However, browsing through the
software I did not find any of them fulfilling my requirements (Newer
versions might include what you are looking for). After spending weeks
trying out and searching for appropriate software, I found that creating
FSTAT inputfiles for each pair of populations and obtaining 95% CI for
the overall Fst was the best (only?) solution. Know its a very tedious
way around the problem, but with <100 pairs, good copy/paste skills and
music on the ipod it should be possible to do in a day or two...
===I encountered the same problem in a study of ours.
I decided to use "Resampling Stats" to calculate a bootstrapped "average
pairwise Fst) across all our AFLP-loci. Maybe this is not the most elegant
way how to do it, but it is at least one way to do it!
===I'm not aware of any software that can perform that task. I've recently
had the same problem and ended up analysing 66 datapools seperately (in
FSTAT) to get the C.I's on pairwise Fst's. Just a couple a weeks ago there
was quite a similar question here on Evoldir concerning C.I.'s on Fis values
on subpopulation-level. It seems to me that there is a big need for a piece
of software to calculate C.I's on all levels. RSTCALC will calculate all the
desired C.I's and tests, but only on Rst values. Guess that wasn't much
help, but if you get some usefull replies, please post the answers.
===I believe TFPGA (free software) will accomplish what you're looking for.
Here is the URL: http://www.marksgeneticsoftware.net/tfpga.htm
TFPGA provides "Theta-P" values, but these are equivalent to Fst.
===You can do this in GDA (Genetic Data Analysis - Paul Lewis -
free download from Paul's Web site). You'll have to select on pair
of populations at the time, although (easily done within the program,
ie you don't have to change the input file). Have you considered using
permutation to test for the significance of your pairwise Fst values? This
can be easily done in Fstat or Genetix.
===I have used GDA
(http://hydrodictyon.eeb.uconn.edu/people/plewis/software.php) for
some of these types of calculations. In a single dataset with multiple
populations, you can 'Exclude' (using the 'Misc.' menu) all populations
but two, then calculate an FST and bootstrap over loci (options within the
F-stats menu) to get a confidence interval. In order to change parameters
of the bootstrap, go to F-stats --> Options --> Options for bootstrap.
Unfortunately, you would have to do this for each pair of populations,
but it seems to work well. Good luck!
===The program GDA (Genetic Data Analysis; Lewis and Zaykin 1999)
will bootstrap over loci to give confidence intervals for pairwise Fst
(theta). You have a be a little tricky to get the estimate though.
Once you input your data into GDA, it will give you the option of
excluding populations from each analysis. You have to exclude all
populations but the two that for which you are interested in getting a
pairwise estimate, and then estimate Fst and bootstrap over loci for
the confidence interval. It won't give confidence intervals for the
estimates in the pairwise theta matrix it estimates. It takes a bit
longer this way, but I don't know of other programs that will give
confidence intervals for pairwise estimates. GDA can be found at :
http://hydrodictyon.eeb.uconn.edu/people/plewis/software.php
philippe.girard@umontreal.ca