v1.2.1

• Don't average likelihood over bootstraps by default.
• Set do_bootstrap flag to false by default for methods performing nested model comparison.

v1.2.0

• Adding dominance coefficient h to DFE inference. h can either be fixed or inferred jointly with the DFE parameters. We can also introduce a relationship between h and S. When estimating h, precomputation over a grid of dominance coefficients takes some time. Implementation was validated with SLiM.
• By default, parameters are now fixed to infer a semidominant (h=0.5) deleterious DFE without correcting for ancestral misidentification (eps=0).
• Improved bootstrapping. By default, 2 runs are carried out per bootstrap sample and the most likely result is taken which previously controlled the number of retries in case of optimization failure). Bootstrapping is now also carried out by default, and mean and standard deviation across bootstraps are logged.
• Initial optimization runs are now recorded
• Added DFE class representing a frozen parametrization.
• Expanded documentation on SFS parsing and DFE inference.
• Allow to specify how the point estimate is determined when plotting discretized DFE with confidence intervals.
• Refactored InferenceResult class.
• Refactored methods returning CIs.

v1.1.13

Fixed bootstrap issue where seeding caused unwanted correlation between the resampled neutral and selected SFS, which could result in smaller confidence intervals.

v1.1.12

Made cyvcf2 optional via [vcf], improved DFE plot labels, added ancestral misidentification support and unnormalized Watterson’s theta to Spectrum class.

v1.1.11

Support for passing alternative optimizer to BaseInference and JointInference.

v1.1.10

Allow for probabilistic polarization when parsing SFS by looking at ancestral allele probability VCF info tag.
Allow the transition/transversion ratio in the K2SubstitutionModel to be fixed to the value observed in the data.
Adjust LRTs to account for parameters near boundaries.
Extend ExistingOutgroupFiltration so that number of missing outgroups can be specified.
Add RandomStratification and ContigFiltration classes.

v1.1.9

Add simulation class for simulating SFS data with known DFE.

v1.1.8

Update cyvcf2 dependency to fix broken wheel for Mac ARM (brentp/cyvcf2#305). Fix minor problem with remote files when disabling file caching.

v1.1.7

Implement serialization of maximum likelihood ancestral annotation to allow for later inspection of results. Improve ancestral allele info tag information for site where annotation was not possible.

v1.1.6

Lazy-load some modules to allow faster initial loading. Update docs.