<div dir="ltr">Hi,<br><div class="gmail_quote"><div dir="ltr"><div class="gmail_quote"><div dir="ltr"><div><br></div><div>We are very happy to announce the release of HDDM 0.5 (hierarchical Bayesian estimation of the drift-diffusion model) which comes with a lot of new features and enhancements. The main reason to upgrade is the new default model that does very well in parameter recovery studies and has much faster convergence.<br>
</div><div><span style="line-height:22.386363983154297px;text-align:justify;font-size:medium;font-family:Verdana,Arial,sans-serif"><br></span></div><div>Installation instructions can be found on our homepage:</div>
<div><a href="http://ski.clps.brown.edu/hddm_docs/" target="_blank">http://ski.clps.brown.edu/hddm_docs/</a><br></div><div><br></div><div>To upgrade from an existing installation, type:<br></div>
<div>pip install -U --no-deps kabuki<br></div><div>pip install -U --no-deps hddm<br></div><div><br></div><div>Changes:</div><div><br></div><div><div>HDDM 0.5</div><div>
========</div><div><br></div><div>* New and improved HDDM model with the following changes:</div><div> * Priors: by default model will use informative priors</div><div> (see <a href="http://ski.clps.brown.edu/hddm_docs/methods.html#hierarchical-drift-diffusion-models-used-in-hddm" target="_blank">http://ski.clps.brown.edu/hddm_docs/methods.html#hierarchical-drift-diffusion-models-used-in-hddm</a>)</div>
<div> If you want uninformative priors, set ``informative=False``.</div><div> * Sampling: This model uses slice sampling which leads to faster</div><div> convergence while being slower to generate an individual</div>
<div> sample. In our experiments, burnin of 20 is often good enough.</div><div> * Inter-trial variablity parameters are only estimated at the</div><div> group level, not for individual subjects.</div><div> * The old model has been renamed to ``HDDMTransformed``.</div>
<div> * HDDMRegression and HDDMStimCoding are also using this model.</div><div>* HDDMRegression takes patsy model specification strings. See</div><div> <a href="http://ski.clps.brown.edu/hddm_docs/howto.html#estimate-a-regression-model" target="_blank">http://ski.clps.brown.edu/hddm_docs/howto.html#estimate-a-regression-model</a> and <a href="http://ski.clps.brown.edu/hddm_docs/tutorial_regression_stimcoding.html#chap-tutorial-hddm-regression" target="_blank">http://ski.clps.brown.edu/hddm_docs/tutorial_regression_stimcoding.html#chap-tutorial-hddm-regression</a></div>
<div>* Improved online documentation at</div><div> <a href="http://ski.clps.brown.edu/hddm_docs" target="_blank">http://ski.clps.brown.edu/hddm_docs</a></div><div>* A new HDDM demo at <a href="http://ski.clps.brown.edu/hddm_docs/demo.html" target="_blank">http://ski.clps.brown.edu/hddm_docs/demo.html</a></div>
<div>* Ratcliff's quantile optimization method for single subjects and</div><div> groups using the ``.optimize()`` method</div><div>* Maximum likelihood optimization.</div><div>* Many bugfixes and better test coverage.</div>
<div>* hddm_fit.py command line utility is depracated.</div></div><div><br></div><div>Best,</div><div>Thomas, Imri and Michael</div></div>
</div><br></div>
</div><br></div>