Change History

Version 1.1.1

Changes

  • Updating documentation configuration, phycoeval tutorial, and installation instructions.

  • Fixing bug introduced in v1.1.0 when allowing sites with no data for some leaf populations. When merging site pattern probabilities from the top of child branche to the bottom parent branches, we were looping over the number of children, rather than the number of children with data, which was causing an indexing error. This release fixes this bug.

Version 1.1.0

Breaking changes

  • Implementing smarter handling of missing data.

    Previously, even if only one leaf population was missing data for a character, we threw out the whole character (mirroring the behavior of beast/SNAPP). However such a character can be informative about the pop sizes and relationships of other populations.

    Now, such a missing character is essentially ignored only for the clade(s) from which it is missing.

    This will be needed for introducing migration (phylo networks), where parent nodes of reticulating nodes can recceive no allele copies (going back in time) from the daughter. So, we need to account for the probability of a node having no allele copies.

    Results of analyses using previous versions that relied on the --relax-missing-sites flag will differ from Version 1.1.0 onward. In previous versions, this option enabled sites with missing data from one or more tip populations to be completely ignored. Now, such sites (unless data are missing from ALL tip populations), will be used and influence the results.

Changes

  • Updating test suite behavior to avoid running slow tests by default.

  • Updating dev tools. Updating build scripts for dependencies and adding test scripts that are more HPC friendly.

  • Adding cladogram ouptut option to sumphycoeval. This makes it easier to see shared/multifurcating divergences.

  • Adding tests to confirm that loci (even individual sites) can be disjoint in an alignment. This should be true, because each site is independent and only ends up being a count of total and “red” alleles (i.e., the sequence labels are ignored). For example, the following two alignments are represented identically as biallelic data by ecoevolity (where “pop-1” and “pop-2” are the population labels):

    individual-1-locus-1_pop-1  ATT???
individual-2-locus-1_pop-1  ATT???
individual-1-locus-1_pop-2  TTT???
individual-2-locus-1_pop-2  ATA???
individual-1-locus-2_pop-1  ???GGC
individual-2-locus-2_pop-1  ???GGC
individual-1-locus-2_pop-2  ???GGA
individual-2-locus-2_pop-2  ???TGC

    and:

    individual-1_pop-1  ATTGGC
individual-2_pop-1  ATTGGC
individual-1_pop-2  TTTGGA
individual-2_pop-2  ATATGC

Version 1.0.0

Changes

  • Adding phycoeval, sumphycoeval, and simphycoeval tools and associated code and tests. These tools implement fully phylogenetic models of shared or multifurcating divergences under a generalized tree distribution.

  • Updating documentation.

  • Documenting the Pitman-Yor and Uniform distributions over divergence models for ecoevolity.

  • Updating nex2yml tool to take both ecoevolity and phycoeval config files.

Version 0.3.2

Changes

  • Fixing bug related to NAN likelihoods. Under extreme combinations of parameters, NAN likelihoods were being logged, and no error or warning occurred. This release fixes that issue. Any analyses prior to this release that did not report NAN likelihoods were not affected by this bug.

Version 0.3.1

Changes

  • Fixing bug in simcoevolity that prevented the ‘–charsets’ and ‘–max-one-variable-site-per-locus’ options from being used together. There was a sanity check from before the ‘–charsets’ option existed that needed to be skipped when the ‘–charsets’ option was used.

Version 0.3.0

Changes

  • Adding new simulation option to simcoevolity. This new option allows simulating multi-locus datasets (each locus comprised of multiple linked sites) that matches the locus lengths and missing data patterns of the empirical dataset. The information to simulate such datasets is provided in a nexus ‘sets’ block that delineates the locus boundaries with ‘charsets’.

Version 0.2.1

Changes

  • Adding default constructor for MatrixExponentiator to appease Clang. This change does not affect behavior; it simply allows older versions of Clang to compile the code.

  • Adding more documentation.

Version 0.2.0

Changes

  • Allowing the MCMC operator ‘TimeRootSizeMixer’ to be specified as a ‘global’ operator (rather than only tree-specific), and making this the default behavior. The default behavior before this release was for each tree (population pair) to have it’s own ‘TimeRootSizeMixer’ operator. Having ‘TimeRootSizeMixer’ as a global operator should improve mixing when pairs share divergence times.

Version 0.1.0

  • Initial release.