This is a BETA version of Hickory. Although we believe the results to be reliable, please treat them with caution and check with us before publishing any results based on analyses performed with this version. Several users of earlier versions encountered errors when analyzing large data sets. The errors arose because of memory allocation problems that weren't properly handled. Large data sets may give similar errors with this version, but it is more memory efficients, so you should be able to handle larger data sets before running into problems. Please let us know of any problems you encounter with this version. The documentation included in this distributions still refers to the old version, but there are only two changes (described below) that you'll need to be aware of. We hope to have a release version (unless severe errors are reported) by the end of February. It will contain updated documentation, including a more complete explanation of the new parameters associated with estimating among-population correlations. Changes in this version ----------------------- 1) Hickory now uses slice sampling for sampling univariate parameters. Slice sampling guarantees that each sample is drawn directly from the full conditional. As a result, the column of acceptance probabilities you've seen in earlier versions has disappeared. There will be a table of acceptance probabilities for any multivariate parameters, i.e., those associated with loci having multiple alleles. This change results in a sampler that runs somewhat more slowly, but autocorrelation in the chain is substantially less. As a result, we've changed the default burnin, sample, and thin parameters from 50,000, 250,000, and 50, respectively to 5000, 100,000, and 20. Be sure to take a look at the trace for f and theta to get a feel for whether the sampling is adequate. If you're paranoid, save the log file, and run some formal diagnostics using boa() in R. 2) You'll see a new option on the analysis menu: "Estimate Theta-Y". It is checked by default. Theta-Y allows us to provide estimates of the among-population correlation as described in Song et al. (Evolution 60:1; 2006). Theta-I, Theta-II, and Theta-III are described in that paper. If you uncheck the option to estimate Theta-Y (or use the --nothetay flag on the command line), you'll get an analysis equivalent to those in previous versions, where Theta-II is equivalent to the old Theta-B. Building from source -------------------- In the unlikely event you want to build Hickory from source, I'd advise waiting until the final distribution. The Makefile is horribly hacked and will require some fairly severe modifications to work on any systems other than ours. The distribution version will come complete with a configure script that will (should) work on any Unix-like system, provided you have the appropriate support libraries installed.