Old News

• ∂a∂i version 1.6.2 released; Dec 4, 2011

This release fixes a long-standing bug in make_fux_table. (Testing suggests that in almost all cases the bug had little effect on results.) Also fixed is a bug that prevented optimizations from succeeding.

• ∂a∂i version 1.6.1 released; Nov 8, 2011

This release improves support for modeling admixture. It turns out that when the admixture methods in PhiManip are used, this can result in oscillations in the calculated frequency spectrum as the grid size is changes, which invalidates extrapolation. One solution is to run with pts_l of length one, so there is no extrapolation. Another solution, if the admixture is the last event in your model, is to implement admixture using the new admix_props argument to Spectrum.from_phi, rather than using the PhiManip methods. * ∂a∂i version 1.6 released; July 8, 2011

This release introduces support for modeling temporal samples, in which samples from different timepoints are compared. This is implemented through freeze arguments to the Integration functions. See the manual under "Ancient Sequences" for more information.

• ∂a∂i version 1.5.3 released; May 25, 2011

This release includes numerous small bug fixes and enhancements. Also, binaries are now provided for Python 2.7.

• ∂a∂i version 1.5.2 released; October 6, 2010

This release reverts to the previously used optimizer, as the bounded version can be much slower if the maximum-likelihood parameters are far from the parameter bounds.

• ∂a∂i version 1.5.1 released; August 6, 2010

The big news for this release is the use of an improved version of the BFGS optimizer, which deals with bounded parameters much better. This should dramatically speed up fits when the optimal parameters are at one or more of the bounds. There are also several minor bugfixes.

The new optimizer is a scipy wrapper of code developed by Ciyou Zhu, Richard Byrd, and Jorge Nocedal. The algorithm is described in RH Byrd, P Lu, J Nocedal SIAM J Sci Stat Comput 16:1190 (1995) and C Zhu, RH Byrd, J Nocedal ACM T Math Software 23:550 (1997).

• ∂a∂i version 1.5.0 released; July 13, 2010

The first major enhancement in this release is the use of a new default grid for integrating the PDE. The new Numerics.exponential_grid was suggested by Simon Gravel. It is more tightly packed toward x=0 and x=1 than the old grid. Additionally, the integration over the PDE result to get the FS has been improved by using a more analytic formulation. As a result, you should get much more accurate results with lower numbers of grid points. Hopefully this will speed up your analyses.

Additionally, methods were added to dadi.Plotting to plot 3D spectra. This includes a method which employs MayaVi2, a remarkable visualization library. If you wish to use this function, you'll need to install MayaVi2, which is most easily accomplished through the Enthought Python Distribution.

• ∂a∂i version 1.4.1 released; April 20, 2010

This release enhances Windows compatibility (thanks to Eduardo Guerra Amorim for testing). An installer for Windows is now provided. Also, a grid parameter search method was added.

• ∂a∂i version 1.4.0 released; March 9, 2010

This release of ∂a∂i includes our first formal manual, plus small changes to improve ease-of-use. The manual is in doc/manual/manual.pdf or it can be accessed from the link on the right. An option was added to the optimization methods to hold particular parameters at fixed values, easing likelihood-ratio testing.

• ∂a∂i version 1.3.4 released; February 9, 2010

This release of ∂a∂i includes two small changes to improve ease-of-use. First, population ids are now stored in the Spectrum objects, and these will be used by default in plotting routines. Second, the model-fitting methods have been updated to support models with extra fixed parameters. For example, one might want to analyze different data sets with different fixed values of theta, and this will help support that usage.

• ∂a∂i paper published; November 6, 2009

PLoS Genetics has published the paper describing ∂a∂i.

Also, I have posted source code for ∂a∂i 1.3.3. This release fixes a minor bug (which did not affect computation results), and hopefully enables Python 2.4 compatibility.

The paper describes the method and infers demographic models for human expansion out of Africa and the settlement of the New World. The out of Africa model is then combined with a previously estimated distribution of selection coefficients to accurately predict the distribution of segregating nonsynonymous variation between populations.

• ∂a∂i version 1.3.0 released; October 22, 2009

This release of ∂a∂i includes two substantial changes to improve ease-of-use. In addition, the tests suite has been expanded. * The Misc.make_data_dict function has been added to ease import of data into ∂a∂i. See DataFormats for details on the file format for this method. * Spectrum objects now track whether or not they are folded. This reduces the potential for mistakes, and it allows operations like projection and marginalization of folded spectra to be handled automatically and correctly.

• Paper accepted!; September 7, 2009

The paper on ∂a∂i has been accepted for publication in PLoS Genetics. It has also been posted on the arXiv.

• Version 1.2.3 release; May 21, 2009

Primarily, this version fixes several small bugs. It also adds an extension of the IM model to Demographics2D, support for pulling specific populations out of an ms file, and a non-log-parameters optimization method.

• Version 1.2.2 release; March 23, 2009

Added method to import frequency spectra from SFS_CODE output. Also added the classic Isolation-Migration model to Demographics2D. Finally, some small bugfixes and additional technical features were added.

• Version 1.2.1 released; February 17, 2009

Added support for Hessian calculation, which can be used to roughly estimate uncertainties. Additionally, crashing bug related to saving spectra to file was fixed.

• Version 1.2.0 released; January 23, 2009

The way in which integration timesteps are assigned was changed. The new method better respects the symmetries in the diffusion equation and yields more stable results in extrapolation. This will require changes to user code. See the YRI_CEU example for hints as to what is needed.

Also, support was added for bias correction in the case in which SNPs were ascertained as heterozygous in sequencing one individual (who is not in the current sample). This allows ∂a∂i to be applied to certain genotyping data sets.

• Version 1.1.0 released; November 20, 2008

Support was added for dominant selection. In addition, ∂a∂i.Misc now includes a method to generate the table of f_{ux} that is needed for applying the ancestral misidentification correction.

• Version 1.0.0 released; November 11, 2008

To the right (under featured downloads) you will find the source for ∂a∂i-1.0.0 and an installer for Mac OS X. Included in the source distribution are a couple of short examples of using ∂a∂i. Also included is API documentation ∂a∂i's functions (also linked to the right). Enjoy!

• Public release soon; October 16, 2008

As we've been presenting our results, people have been asking if they can use ∂a∂i on their own data. Rest assured, ∂a∂i will be made freely available in a form that you can use. We're not, however, quite there yet. In particular, useful documentation is forthcoming. We expect to have it up (and ∂a∂i's first official release) by the middle of November. If you subscribe to the dadi-announce mailing list (see link to right), you'll be informed by email when ∂a∂i is officially released and upon major updates.