https://r-forge.r-project.org/ R-Forge Project News Highlights https://r-forge.r-project.org/forum/forum.php?forum_id=4533 The 1.4 version of depmixS4 has an important (and much requested) new feature: the possibility to request standard errors of estimated parameters through the use of a finite differences approximation of the hessian. Comments and feedback much appreciated! After some time in this beta version we will post the new version to CRAN as well. <br /> Ingmar Visser &amp; Maarten Speekenbrink depmixS4 - hidden Markov model classes Ingmar Visser 2019-01-14T07:17:02+00:00 https://r-forge.r-project.org/forum/forum.php?forum_id=4424 depmixS4 version 1.3-0 has been released on CRAN. See the NEWS file<br /> for an overview of all changes. The most important user-visible changes are:<br /> <br /> 1) more compact pretty-printing of parameters in print/summary of<br /> (dep)mix objects (following lm/glm style of presenting results)<br /> <br /> 2) some speed improvements in the EM algorithm, most notable in large data/models<br /> <br /> 3) EM has an optional argument to use the classification likelihood<br /> instead of the usual likelihood; this can be useful as a means of<br /> starting value generation; use with caution as results are often<br /> unstable.<br /> <br /> Best, happy mixing, Ingmar depmixS4 - hidden Markov model classes Ingmar Visser 2013-09-20T22:33:27+00:00 https://r-forge.r-project.org/forum/forum.php?forum_id=4423 depmixS4 version 1.3-0 has been released on CRAN. See the NEWS file<br /> for an overview of all changes. The most important user-visible changes are:<br /> <br /> 1) more compact pretty-printing of parameters in print/summary of<br /> (dep)mix objects (following lm/glm style of presenting results)<br /> <br /> 2) some speed improvements in the EM algorithm, most notable in large data/models<br /> <br /> 3) EM has an optional argument to use the classification likelihood<br /> instead of the usual likelihood; this can be useful as a means of<br /> starting value generation; use with caution as results are often<br /> unstable.<br /> <br /> Best, happy mixing, Ingmar depmixS4 - hidden Markov model classes Ingmar Visser 2013-09-19T09:28:03+00:00 https://r-forge.r-project.org/forum/forum.php?forum_id=4336 Version 1.2-0 has been released on CRAN. The major change in this version is that it supports missing data in the responses (missing covariates result in an error). The vignette accompanying the package has a brief explanation of how missing data is treated within the EM algorithm and using direct optimization of the loglikelihood. depmixS4 - hidden Markov model classes Ingmar Visser 2012-07-04T16:52:15+00:00 https://r-forge.r-project.org/forum/forum.php?forum_id=4054 The main change in this version is that the main loop for computing the forward and backward variables has been moved into C resulting in considerable speed improvements. Depending on model specifics this results in a 2-4 fold speed increase when fitting models.<br /> <br /> Another noteworthy change is the addition of several examples such as a Poisson change point model (with artificial data, see ?responses), and an example of a financial time series (logged returns from the S&P 500, thanks to Chen Haibo for providing the example, see ?depmix). depmixS4 - hidden Markov model classes Ingmar Visser 2011-11-14T19:20:23+00:00 https://r-forge.r-project.org/forum/forum.php?forum_id=3062 A new version of depmixS4 is submitted to CRAN which fixes a number of bugs. One related to multinomial observation models with n&gt;1 and a logit link function, and one concerning the generation of starting values. depmixS4 - hidden Markov model classes Ingmar Visser 2010-11-25T15:15:00+00:00 https://r-forge.r-project.org/forum/forum.php?forum_id=2876 depmixS4 has reached some form of maturity and therefore we have bumped its version number to 1.0-0. The reason for this is that the paper illustrating the possibilities of depmixS4 has now been published in the Journal of Statistical Software and can be found there: http://www.jstatsoft.org/v36/i07<br /> <br /> Except for a few minor issues there are no large changes since the previous version except the addition of a vignette which has the same contents as the JSS paper. depmixS4 - hidden Markov model classes Ingmar Visser 2010-09-17T12:26:50+00:00 https://r-forge.r-project.org/forum/forum.php?forum_id=2254 The 0.3-0 version of depmixS4 contains an example of a model with multivariate normal distributions on the ?makeDepmix help page and some minor improvements in code and documentation. depmixS4 - hidden Markov model classes Ingmar Visser 2010-02-22T09:10:43+00:00 https://r-forge.r-project.org/forum/forum.php?forum_id=1530 This new release mostly has some minor changes and an important bug fix in multinomial response models with covariates (giving odd results and violating the baseline constraint in baseline multinomial logit models). depmixS4 - hidden Markov model classes Ingmar Visser 2009-07-01T21:36:30+00:00