Updated package version and date in DESCRIPTION and website documentation to 3.3.0.03.
Updated logo path in README.md to point to man/figures/logo.png for consistency with package structure.
Added favicon and manifest links to HTML documentation files for improved branding and browser integration.
Added the package website (https://hreinwald.github.io/drc) as the primary URL in the DESCRIPTION file for better discoverability.
Added the rss() function to the reference index in _pkgdown.yml.
Added logo image to the dose-response workflow vignette and updated the vignette date.
Simplified labeling of effective dose (ED) estimates in the workflow vignette outputs for clarity, removing the e:1: prefix.
Updated model comparison output in the vignette to include additional columns and more precise values.
drc 3.3.0.02
New Features
Added rss() function for computing the residual sum of squares of a fitted drc model. Refactored Rsq() to reuse rss() internally; both functions are now exported.
Bug Fixes
Fixed ED() for exponential decay models (EXD.2, EXD.3, AR.2, AR.3, W1.x, W2.x) with two fixed parameters: when only one parameter is estimated (1×1 variance-covariance matrix), the function previously failed with “incorrect number of dimensions” errors. Enhanced ED.drc to defensively coerce scalar/vector vcov inputs to proper matrices and to always strip names from gradients for consistent matrix algebra. This fix now allows retrieving ED values from exponential decay models with two fixed parameters, which was previously impossible.
Fixed gradient handling in ED() to ensure model-specific derivative functions always return unnamed numeric vectors, preventing dimension errors in delta-method standard error calculations.
Fixed boundary detection bugs in MAX(): used unname() so named return values from cedergreen models are compared correctly with unnamed lower/upper scalars, and added tolerance in boundary check since numerical optimizers return values near but not exactly at boundaries.
Fixed PR() dropping ... arguments for single-curve models.
Fixed all 17 issues in ucedergreen() function: missing +c term in model formula, edfct signature mismatch with the drc framework, undefined xlogx function call in deriv1, missing match.arg() validation for method, vectorized | operators in scalar if() guards, missing useFixed flag computation, maxfct signature mismatch and unsafe parameter indexing, broken self-starter ignoring alpha/method/useFixed, missing fctName/fctText parameters, deriv1 excluded from return list, and documentation issues.
Fixed SE calculation for absolute type ED(): the model-specific edfct gradient functions treated asymptote parameters as constants when type="absolute", missing the chain-rule contribution from the absToRel conversion and underestimating the standard error. Now uses numerical central differences with an improved adaptive step size. Added internal helpers .centralDiffGradient(), .safeConfintBasic(), and .computeSE() to make SE computation more robust: .computeSE() guards against non-positive-definite variance-covariance matrix slices (returning NA instead of erroring), and .safeConfintBasic() validates residual degrees of freedom before calling confint.basic(), falling back to a z-distribution when df.residual() returns an invalid value.
Fixed inverted otrace/silentVal logic in drmOpt() where otrace=TRUE incorrectly caused silent=TRUE in try(optim()), suppressing error messages instead of displaying them.
Fixed searchdrc() regex error and convergence failure behavior.
Fixed citation URL: reordered URLs in DESCRIPTION so citation('drc') returns the GitHub repository URL instead of r-project.org.
Fixed ED() “incorrect number of dimensions” error for models with few estimated parameters (e.g., EXD.3 with fixed c and d): ensured indexMat is always treated as a matrix before column subsetting.
Fixed ED() returning NaN with warning for LL.5 models with ill-conditioned parameters: added validity check to return Inf (indicating EC50 is outside valid range) instead of NaN when exp(-tempVal/parmVec[5]) - 1 is non-positive. Also fixed NaN handling in the check condition to prevent “missing value where TRUE/FALSE needed” errors in backfit() and other functions.
Fixed additional robustness issues in ED() / ED.drc: loop now always iterates over all curves and all response levels, filtering by clevel after computation rather than before; invMatList is grown dynamically to avoid NULL holes; curve label construction uses a single structured object with explicit match and display fields; variance-covariance matrix slices always use drop = FALSE to remain matrices.
Fixed mselect() missing two closing braces that caused a parse error when the function was sourced directly.
Fixed ED.lin.R bugs: removed a duplicate if-block (dead code that evaluated the same condition twice), removed a stray debug print() statement, and added the missing parameterNames = c("b0", "b1", "b2") argument to the deltaMethod() call for quadratic models (the omission caused incorrect parameter mapping and wrong confidence intervals).
Fixed gammadr() first-derivative (deriv1) calculation: the gradient with respect to the dose parameter incorrectly used parmMat[, 1] (the rate parameter) where dose was required, producing wrong gradient values.
Fixed maED() model-averaging: models whose ED estimates are non-finite (Inf or NaN) are now detected and excluded from the weighted average (with a warning naming the model and the offending values); models that returned a try-error during fitting are also excluded. When all candidate models are excluded, the function returns NA for all estimates instead of 0 or NaN.
Added warning to noEffect() when degrees of freedom difference is ≤ 0, clarifying that the likelihood ratio test may not be meaningful when the dose-response model has no additional parameters compared to the null model (e.g., when most parameters are fixed).
Changes
Added NEWS.md version control log. Reformatted legacy news file into properly formatted NEWS.md with categorized sections.
Improved documentation for Weibull starting value method parameter across weibull1(), weibull2(), and all wrapper functions (W1.2, W1.3, W1.4, W2.2, W2.3, W2.4, AR.2, AR.3, EXD.2, EXD.3).
Enhanced roxygen2 documentation for ED and ED.drc functions with improved parameter descriptions and examples.
Added comprehensive test suites for anova.drclist,summary.drc, print.summary.drc, noEffect, searchdrc, backfit, getInitial, drmEMeventtime, repChar, rdrm, gompertzd, MAX(), and PR() functions.
Added comprehensive test suites for llogistic/LL.x models, weibull1/W1.x/EXD.x models, logistic.ssf, gammadr, EDcomp, mselect, drmOpt, modelFunction, modelFit, anova.drclist, rss, and ED.lin.
Large-scale dead code removal across 70+ R source files: removed commented-out function implementations, stray print() debug statements, old code paths, and if(FALSE){...} blocks. No logic changes; all roxygen2 documentation and meaningful explanatory comments were preserved.
Removed dead code iband.R and all associated references.
Removed unused inst/citation file, superseded by CITATION.cff at repository root.
Deleted build_pkgdown.R build script.
Added PLoS ONE 2015 article and CRC Press 2019 book references to CITATION.cff.
Updated installation instructions and README documentation.
Added magic to Suggests in DESCRIPTION for test dependency.
drc 3.3.0.01
New Features
Created comprehensive vignettes: dose-response-workflow.Rmd providing a complete tutorial on dose-response analysis, and nec-models.Rmd documenting No Effect Concentration modeling with NEC.2/NEC.3/NEC.4 function variants.
Set up pkgdown website infrastructure: added _pkgdown.yml with Bootstrap 5 configuration, created build_pkgdown.R script for build automation, documented pkgdown build process in README, and generated pkgdown documentation site.
Added computationally robust (stable) wrapper functions in new ED_robust.R module: ED_robust() for calculating ED values with proper error handling that returns NA instead of failing when an ED value is not estimable, maED_robust() for model-averaged ED estimation with the same graceful error handling, and get_ed_interval() for recommending appropriate confidence interval methods based on model type.
Added comprehensive test suite covering ED calculations, predictions, plotting, residuals, model selection, and utility functions.
Added drm_name() helper function to ED_robust.R.
Enhanced package startup message with citations and developer credits.
Added drm_legacy() as an internal reference function preserving the original drm() implementation.
Added testthat infrastructure with tests verifying drm() output matches drm_legacy() output across continuous, binomial, Poisson, and negative binomial data types.
Added comprehensive anova tests.
Bug Fixes
Fixed vignette build by removing vignettes from .Rbuildignore and correcting incorrect mselect() usage in examples.
Fixed Rd comment warning by escaping the %*% operator in documentation.
Fixed all devtools::check() errors and warnings: added roxygen2 @keywords and lifecycle deprecation notices for deprecated CRS functions, expanded dataset documentation files with examples, added missing dataset aliases, and fixed Weibull model documentation.
Fixed unsafe global state modification via options(warn), incorrect compParm od/pool handling, and residuals division by zero.
Fixed NaN warning in summary.drc for robust estimation methods (metric trimming, Winsorizing, Tukey’s biweight).
Improved predict.drc and vcov.drc to resolve 23 test failures.
Fixed mselect() to always compute Lack of fit p-values for all models, not only when nested=TRUE.
Fixed a bug in anova.drclist where negative or non-finite F statistics produced NaN p-values; negative F statistics now return p-value of 1 and non-finite F statistics return NA.
Fixed duplicate aliases and unstated dependencies in examples.
Fixed package dependency warnings: added data.table and dplyr to Imports, updated NAMESPACE with required imports.
Fixed S3 method consistency issues and changed confint.basic roxygen tag from @exportS3Method to @export.
Fixed escaped LaTeX special characters in roxygen2 documentation and Rd files.
Completed comprehensive roxygen2 documentation audit: added missing @param tags, removed dontrun/donttest wrappers to enable automated example testing, and fixed broken examples across documentation files.
Enhanced dataset documentation: improved descriptions and fixed typos in dataset .Rd files, added examples sections to dataset .Rd files that were missing them.
Improved confint.drc robustness: added stop() fallback to switch() in confint.basic() to handle unknown intType values gracefully instead of returning silent NULL.
Removed @export from confint.basic() as internal helpers should not be part of the public API.
Enhanced roxygen2 documentation for CRS.5, convenience functions, and ED_robust with improved argument descriptions.
Updated DESCRIPTION: added Hannes Reinwald as maintainer and co-author, updated package version to 3.3.0.01.
Removed external drcData package dependency; example datasets are now bundled directly in the package data/ directory.
Added .Rd documentation files for all bundled datasets.
Renamed internal variables for clarity: ndRows to nRows in predict.drc, posIdx to validVar in summary.drc.
Added test coverage documentation.
Renamed all 33 R source files from lowercase .r to uppercase .R extensions for consistency.
Updated all .Rd documentation files to reference the new file names.
Added GNU General Public License version 2 file and updated license version from GPL-2 to GPL-2.0 in DESCRIPTION.
Added Hannes Reinwald as author in the DESCRIPTION file.
Updated README with revised installation instructions and bug report link.
Migrated all package documentation to roxygen2-generated Rd files.
Regenerated NAMESPACE via roxygen2.
Added @exportS3Method tags to S3 methods in confint.drc.R and mrdrm.r.
Updated package version format from 3.3-0 to 3.3.0.
Lowered the minimum R version requirement to 4.0.0.
Added new CRS.5 wrapper function and CRS.6 six-parameter model where the alpha exponent is estimated rather than fixed.
Bug Fixes
Fixed a bug where the stop() call for using separate curves with control measurements was inside the if(!noMessage) block, meaning it would be silently skipped when messages were suppressed.
Fixed a bug in noEffect.R where the Poisson null model incorrectly referenced resp instead of using the response vector from the fitted object.
Changes
Refactored the Cedergreen-Ritz-Streibig hormesis model: extracted edfct and maxfct into standalone helper functions (cedergreen_edfct, cedergreen_maxfct), refactored the self-starter function, and improved documentation.
Cleaned up drm() function by removing approximately 900 lines of commented-out dead code, debug print statements, and old experimental implementations.
Removed unused variable isfi and redundant variable lenData (identical to numObs).
Removed a dead loop over pmodelsList2 that could never execute.
Added roxygen2 documentation headers to all R source files across the package.
Fixed typos in source code and manual pages: ‘insted’ to ‘instead’ in gaussian.r and lgaussian.R, duplicate parameter name ‘e1’ to ‘e2’ in ursa.r, ‘contain’ to ‘contents’ in EDcomp.R, ‘mising’ to ‘missing’ and ‘reponses’ to ‘responses’ in drm.Rd, and ‘reponse’ to ‘response’ in CRS.5a.Rd.
Updated DESCRIPTION file: added Encoding field (UTF-8), fixed Authors@R to use proper person() format, removed deprecated Maintainer and LazyLoad fields, added missing Imports (graphics, utils), and updated URLs from HTTP to HTTPS.
Comprehensive repository cleanup and code quality improvements.
