Changes in version 1.1.0 - Fixing CRAN errors. - Added standalone ratioReg() support for weighted ratio regression, optional one-inflation model selection, bootstrap intervals, and diagnostic plots. - Added the oichao() family, a modified Chao estimator based on counts 2 and 3 for improved robustness to one-inflation. - Updated rare-count-family simulation support so oichao() works correctly through the public simulate() methods. - Updated README and package documentation to describe the new oichao() model. - ratioReg(): fixed subset so that bare expressions like subset = grp == "a" evaluate inside data as documented. - ratioReg(): fixed bootstrap with model = "M0" or model = "M1", which previously errored because the primary population estimate was not yet attached when resampling started. - Expanded test coverage for oichao() (mu, variance, density, deviance residuals, analytic population variance) and for ratioReg() (validation paths, maxCount, subset, alternative estimators, extractAIC/nobs, custom ratioFormula, fixed-model bootstrap). - Split family tests into dedicated test_fam_* files and added focused family files for chao() and oichao() to make CI failures identify the affected model directly. - Cyprian Jurkowski contributed the first implementations of oichao() and ratioReg(). Changes in version 1.0.1 - fixed checks of the variables Changes in version 1.0.0 (2026-01-21) - Fixed mixed up ztoi* and oizt* families - Fixed documentation for the estimatePopsize() function - Fixed vignette - Fixed README.Rmd - Logo added [thanks to Oliwia Awuku] - A CITATION file added with reference to the JSS paper Changes in version 0.2.3 (2025-02-13) - BugFix for anova and other not working methods - Tests for not working methods added - Small changes to the documentation Changes in version 0.2.2 (2025-02-04) - Fixing documentation typos and errors - plot now has a default value of qq - Vignette added based on the paper submitted to the JSS Changes in version 0.2.1.2 (2024-07-18) - Bugfix for interaction terms in formula not being considered - Small changes in summary for marginal count distributions - Small fixes for standard errors in predicted means - Code coverage raised to nearly 90% - The logLik.singleRStaticCountData method now has a type argument that, if set to "function" makes the function return the minus log-likelihood function (by default) or if deriv argument is set to 1 or 2 respectively either a gradient or hessian of log-likelihood function. Changes in version 0.2.1.1 (2023-10-23) - Bugfix for tests failing with noLongDouble Changes in version 0.2.1 (2023-10-14) - Fixed bugs in IRLS fitting when providing weights argument when calling estimatePopsize - The weightsAsCounts option in controlModel now works properly, dfbeta and dfpopsize decrease weight of selected row in a model matrix instead of deleting it when this is set to TRUE - simulate method now works for both family object (like ztpoisson()) and for objects returned by estimatePopsize - Introduced singleRStaticCountData sub class for singleRclass and made estimatePopsize a method so that a new package singleRcaptureExtra (under development) can make all necessary calculations for pop size estimation when providing object fitted by countreg::zerotrunc or VGAM::vglm/VGAM::vgam - Some bugfixes for multicore bootstrap - Code was re-factored to make further development/maintenance for the package much easier - Update will be uploaded to CRAN - semiparametric bootstrap now has a much faster sampling algorithm (that does the same job) Unit tests: * Reduced computational burden of unit tests * Multicore tests will only be performed after TEST_SINGLERCAPTURE_MULTICORE_DEVELOPER is set to "true" via Sys.setenv and _R_CHECK_LIMIT_CORES_ to false Changes in version 0.2.0.1 - Added offset argument to estimatePopsize - Added options for parallel computing in bootstrap and in dfbeta - Added deviance for all negative binomial based models. (NOTE: They are very slow for now and I believe it may change after I verify one theoretical results that will lead to significant speed increase for these computations) - Overhaul of starting points (new methods and added linear predictors start in IRLS) - Code for weights in IRLS fitting was speed up - Minor bugfixes Changes in version 0.2.0 (2023-06-11) The package is now at CRAN - features and improvements: - Added final Hurdleztnegbin model - Vastly improved redoPopSize which now handles bootstrap on a fitted model non standard covariance matrixes newdata argument user supplied coef etc. - Added predict.singleR method which offers standard error for both link, response as well as mean predictions - No unexpected warnings should occur now in main function when using the package correctly - All control arguments are now verified before being passed - Fitting is now more reliable - Added information about stats::optim error codes - Added warnings for functions computing deviance - bugfixes: - fixed bugs occurring when using mathematical functions as part of formulas i.e. when setting formula to something like: y ~ log(x) + I(x ^ t) + I(t ^ 2) Changes in version 0.1.4 - features - Added ztoinegbin, oiztnegbin and ztHurdlenegbin models - Added an optional arguments to all family-functions to specify a link function for distribution parameters - Updated and standardized documentation - Added more warnings - Added some more methods for singleR class in some commonly used glm functions, in particular texreg::screenreg should work well now - changes - Changed some default arguments - Added option to save logs from IRLS fitting - bugfixes - Fixed some issues with intercept only models - Fixed some slight miscalculations in information matrixes for one inflated models making fitting them much more reliable - github repository - More and better Rcmd tests Changes in version 0.1.3.2 - features: - Added function that implements population size estimates for stratas - More warnings in fitting - More options in control functions - Corrected/implemented deviance residuals for more models - changes: - Now the whole package uses camelCase - Performance upgrades - Corrected some miss calculated moments - Change exported data so that factors are actually factors not just characters - Removed unused dependency - github repository - Added automated R-cmd check Changes in version 0.1.3.1 - features: - Basically all of documentation was redone and now features most of important theory on SSCR methods and some information on (v)glms - Added checks on positivity of working weights matrixes to stabilise "IRLS" algorithm - Added most of sandwich capabilities to the package, in particular: - S3 method for vcovHC was implemented - vcovCL should work on singleR class objects should work with "HC0" and "HC1" type argument values - Basic version of function redoPopEstimation for updating the population size estimation after post-hoc procedures was implemented - popSizeEst function for extracting population size estimation results was implemented - Minor improvements to memory usage were made and computation was speed up a little - Changed names of mle and robust fitting methods to optim and IRLS respectively - Some bugfixes - More warnings messages in estimate_popsize.fit Changes in version 0.1.3 - features: - Multiple new models - IRLS generalised for distributions with multiple parameters - bugfixes - QOL improvements - extended bootstrap and most other methods for new models Changes in version 0.1.2 - features: - control parameters for model - control parameters for regression in bootstrap sampling - leave one out diagnostics for popsize and regression parameters (dfbetas were corrected) - fixes for Goodness of fit tests in zero one truncated models - computational improvements in IRLS - other small bugfixes Changes in version 0.1.1 - bug fixes and some of the promised features for 0.2.0 in particular - More tiny tests - Some fixes for marginal frequencies - Deviance implemented - dfbetas and levarage matrix - Parametric bootstraps work correctly for the most part there is just some polishing left to do Changes in version 0.1.0 - first version of the package released