{npmlreg} - Record of changes and updates

Version Date
0.43 03/09/07 Uploaded onto CRAN. Otherwise largely identical to version 0.42-1.
0.41 / 0.42-1 08/08/07 Technical improvements concerning the handling of S3 methods and the name space. Random effect standard deviation now available in output. Updated vignette. (not published online).
0.40 29/11/06 Uploaded onto CRAN. Otherwise largely identical to version 0.39-2.
0.39 / 0.39-2 13/11/06 support for up to 600 mass points, minor bug fixes relating the likelihood spike protection and the function post() (not published online).
0.38 04/09/06 Uploaded onto CRAN. Largely identical to version 0.37.
0.37 14/08/06 New function post() for posterior model information, replacing masspoint.classifier() .
0.36-3 / 0.36-4 11/08/06 support for up to 30 mass points, easier access of standard errors via summary()$coef (not published online).
0.36-2 04/08/06 A bug with the handling of weights and offsets in tolfind was fixed.
0.36-1 04/08/06 A bug in the calculation of standard errors for Poisson and Binomial models was fixed. Option vc in tolfind abolished.
0.36 01/08/06 Several adjustments, particularly for the case k=1 .
0.35-8 26/07/06 The necessity to specify the data frame within the weights or offset argument is abolished.
0.35-2 / 0.35-7 25/07/06 Working versions, with several minor improvements and bug fixes (not published online).
0.35 / 0.35-1 12/07/06 Posterior intercept in $post.int; Name Space; object components rearranged (information about the last GLM fit removed and replaced by the correct information concerning the random effect model). All generics as fitted() refer then to the random effect model, too (not published online).
0.34-1 26/06/06 'data' variable robustified against misspecification (not published online).
0.34 06/06/06 First version published on CRAN.
0.33 31/05/06 Improvements with tolfind, in particular with the graphical output.
0.32-2 29/05/06 Support for arbitrary number of masspoints k < 22. tolfind also supported for gq.
0.32-1 23/05/06 Several corrections in manual, help pages, and vignette. sdev is also calculated for Gamma-models and used for starting point selection.
0.32 26/04/06 New data set irlsuicide included. The handbook is now included in form of a vignette.
0.31 23/03/06 Additional option sdev fixing the mixture standard deviation. Several problems with the weight functions were fixed.
0.30 13/03/06 First 'real' package version. Otherwise largely identical to source code version 0.22-2.
0.22-2 08/03/06 Several minor improvements, e.g. an invalid setting of k is 'caught' and gives an appropriate error message.
0.22-1 01/03/06 Names attribute attached to components ebp, post.prob, offset and to the output of predict(). A problem with the predict function for variance component models with random coefficients was fixed.
0.22 27/02/06 Unequal component dispersion parameters also available for allvc. In preparation of inclusion into a formal package, several modifications in order to ensure compatibility of generic functions etc.
0.21 17/02/06 Improved stability and error handling at several occasions; support for the generic function model.matrix() .
0.20-8 15/02/06 Further robustness problems with allvc reported by R. Darnell were fixed.
0.20-7 13/02/06 As the 'glm' class membership had disturbed the update() functionalities, objects are not any more of class 'glm'. In turn, explicit formulation of the family.glmm...() functions.
0.20-6 13/02/06 Several robustness problems with alldist and allvc reported by F. Lawrence and R. Darnell were fixed.
0.20-5 07/02/06 Binomial models can now generally be specified by a matrix of frequencies or by a vector with proportions and a vector of weights. Support for missing values in allvc and tolfind .
0.20-4 15/01/06 Support for factor-valued Bernoulli response, and for cluster ID's other than 1,2,... in allvc .
0.20-3 13/01/06 A problem with Bernoulli data in alldist (with versions 0.20-1 and 0.20-2), caused by the implementation of the missing value routine, was fixed.
0.20-2 11/01/06 A problem with the output of allvc (with versions 0.20 and 0.20-1) was fixed.
0.20-1 06/01/06 Support for missing values in alldist; spike protection for gamma-distributed response in alldist; new output component $model giving the extended design matrix.
0.20 23/12/05 Functionality of function gamdist integrated into alldist; Gamma-distributed response also available for allvc; several changes in output and summaries; new functions plot.glmmNPML and plot.glmmGQ; possibility to specify EM starting mass points.
0.18 08/12/05 All objects are now of class c('glmmNPML/GQ', 'glm'). Additional output $data is given. Support for various standard R functions as e.g. family().
0.17 24/11/05 EM trajectory plots now also available for random coefficients; prediction functionalities for GQ extended.
0.16-4 14/11/05 Generic function update can be applied on all "glmmNPML" and "glmmGQ" objects.
0.16-3 07/11/05 New functions predict.glmmNPML and predict.glmmGQ .
0.16-2 27/10/05 Prior weights are now available for all functions; new option damp.power; summary and graphical output for gamdist improved.
0.16-1 24/10/05 Option verbose added to all functions.
0.16 10/10/05 New functions gamdist and tolfind.
0.15-1 06/10/05 A problem with the likelihood spike protection for negative mass points was fixed. Graphical outputs for non-natural link functions improved.
0.15 05/10/05 Another problem for variance component models with varying cluster sizes, involving the construction of masses from posterior probabilities, was fixed. In addition, the possibility of plotting the disparity over iteration number is provided.
0.14-3 29/09/05 A problem for variance component models with varying cluster sizes, reported by E. Läärä, was fixed.
0.14-2 26/09/05 Protection against likelihood spikes implemented.
0.14-1 19/09/05 Improved graphical output for binomial models.
0.14 30/08/05 EM damping can now be switched on or off
0.13 15/07/05 Additionally to the disparities, also deviances can be calculated now.
0.12 06/07/05 Inclusion of an offset. Additionally, a problem with Bernoulli data reported by R. Darnell was fixed.
0.11 25/06/05 First version of allvc.
0.10 12/05/05 First version of alldist.

[Package npmlreg ]