Merge pull request #17 from ikosmidis/develop

Develop

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: brglm2
 Title: Bias Reduction in Generalized Linear Models
-Version: 0.6.1
+Version: 0.6.2
 Authors@R: c(person(given = "Ioannis", family = "Kosmidis", role = c("aut", "cre"), email = "ioannis.kosmidis@warwick.ac.uk", comment = c(ORCID = "0000-0003-1556-0302")),
 	     person(given = "Kjell", family = "Konis", role = "ctb", email = "kjell.konis@me.com"),
 	     person(given = c("Euloge", "Clovis"), family = c("Kenne Pagui"), role = "ctb", email = "kenne@stat.unipd.it"),

NEWS.md

@@ -1,113 +1,158 @@
-# brglm 0.6.1
+# brglm2 0.6.2
+
+## Other improvements, updates and additions
+* `vcov.brglmFit` objects now uses `vcov.summary.glm` and supports the
+  `complete` argument for controlling whether the variance covariance
+  matrix should include rows and columns for aliased parameters.
+* Deprecated `detect_sepration` and `check_infinite_estimates`, which
+  will be removed from **brglm2** at version 0.8. New versions of
+  `detect_sepration` and `check_infinite_estimates` are now maintained
+  in the
+  [**detectseparation**](https://cran.r-project.org/package=detectseparation)
+  R package.
+* Fixed typos in NEWS.
+
+# brglm2 0.6.1
 ## Bug fixes
 * Fixed bug in AIC reported by `print.summary` for `brmultinom` and `bracl`
-* `detect_separation` now handles one-column model matrices correctly
+* `detect_separation` now handles one-column model matrices correctly.
 
 ## Other improvements, updates and additions
 * Documentation improvements and typo fixes
 
-# brglm 0.6
+# brglm2 0.6
 ## New functionality
-* `brglmFit` can now do maximum penalized likelihood with powers of the Jeffreys prior as penalty (`type = "MPL_Jeffreys`) for all supported generalized linear models. See `brglmControl` and `brglmFit` for details.
+* `brglmFit` can now do maximum penalized likelihood with powers of
+  the Jeffreys prior as penalty (`type = "MPL_Jeffreys`) for all
+  supported generalized linear models. See `brglmControl` and
+  `brglmFit` for details.
 
 ## Other improvements, updates and additions
-* Documentation updates and improvements
-* Updated vignettes to include maximum penalized likelihood with powers of the Jeffreys prior as penalty
-* New examples in `?brglmFit`
+* Documentation updates and improvements.
+* Updated vignettes to include maximum penalized likelihood with
+  powers of the Jeffreys prior as penalty.
+* New examples in `?brglmFit`.
 
-# brglm 0.5.2
+# brglm2 0.5.2
 ## Bug fixes
-* `print.brmultinom` is now exported, so `bracl` and `brmultinom` fits print correctly
+* `print.brmultinom` is now exported, so `bracl` and `brmultinom` fits
+  print correctly.
 
 ## New functionality
-* Added `response_adjustment` argument in `brglmControl` to allow for more fine-tuning of the starting values when `brglmFit` is called with `start = NULL`
+* Added `response_adjustment` argument in `brglmControl` to allow for
+  more fine-tuning of the starting values when `brglmFit` is called
+  with `start = NULL`.
 
 ## Other improvements, updates and additions
-* Documentation updates and improvements
-* Added Kosmidis et al (2019) in the description file
-* Added tests for `brglmControl`
+* Documentation updates and improvements.
+* Added Kosmidis et al (2019) in the description file.
+* Added tests for `brglmControl`.
 
-# brglm 0.5.1
+# brglm2 0.5.1
 
 ## Other improvements, updates and additions
-* Fixed typos in vignettes and documentation
-* Added ORCHID for Ioannis Kosmidis in DESCRIPTION
+* Fixed typos in vignettes and documentation.
+* Added ORCHID for Ioannis Kosmidis in DESCRIPTION.
 
-# brglm 0.5.0
+# brglm2 0.5.0
 ## Bug fixes
-* `brglmFit` now works as expected with custom link functions (mean and median bias reduction)
-* `brglmFit` respects the specification of the transformation argument in `brglmControl`
-* Fixed bug in the computation of the QR decomposition under aliasing in `brglmFit`
-* Other minor bug fixes and performance improvements
-* Protection against use of `quasi`, `quasibinomial` and `quasibinomial` families and documentation update.
+* `brglmFit` now works as expected with custom link functions (mean
+  and median bias reduction).
+* `brglmFit` respects the specification of the transformation argument
+  in `brglmControl`.
+* Fixed bug in the computation of the QR decomposition under aliasing
+  in `brglmFit`.
+* Other minor bug fixes and performance improvements.
+* Protection against use of `quasi`, `quasibinomial` and
+  `quasibinomial` families and documentation update.
 
 ## New functionality
-* Added `bracl` for fitting adjacent category logit models for ordinal responses using maximum likelihood, mean bias reduction, and median bias reduction and associated methods (`logLik`, `summary` and so on)
-* Added `predict` methods for `brmultinom` and `bracl`
-@ Added `residuals` methods for `brmultinom` and `bracl` (residuals of the equivalent Poisson log-linear model)
-* Added the `mis` link functions for accounting for misclassification in binomial response models (Neuhaus, 1999, Biometrika)
+* Added `bracl` for fitting adjacent category logit models for ordinal
+  responses using maximum likelihood, mean bias reduction, and median
+  bias reduction and associated methods (`logLik`, `summary` and so
+  on).
+* Added `predict` methods for `brmultinom` and `bracl`.  @ Added
+`residuals` methods for `brmultinom` and `bracl` (residuals of the
+equivalent Poisson log-linear model)
+* Added the `mis` link functions for accounting for misclassification
+  in binomial response models (Neuhaus, 1999, Biometrika).
 
 ## Other improvements, updates and additions
-* Improved `summary` method for `brmultinom` objects
-* Better starting values for null fits
-* Added references to [arxiv:1804.04085](https://arxiv.org/abs/1804.04085) in documentation
-* Updated reference to [Kenne Pagui et al (2017)](https://doi.org/10.1093/biomet/asx046)
+* Improved `summary` method for `brmultinom` objects.
+* Better starting values for null fits.
+* Added references to [arxiv:1804.04085](https://arxiv.org/abs/1804.04085) in
+  documentation.
+* Updated reference to [Kenne Pagui et al (2017)](https://doi.org/10.1093/biomet/asx046).
 
-# brglm 0.1.8
+# brglm2 0.1.8
 ## Other improvements, updates and additions
-* Improved documentation examples
-* Removed warning about observations with non-positive weights from brmultinom
-* Updated email address for Ioannis Kosmidis in brglmFit
+* Improved documentation examples.
+* Removed warning about observations with non-positive weights from brmultinom.
+* Updated email address for Ioannis Kosmidis in brglmFit.
 
 ## Bug fixes
-* brmultinom returns a fitted values matrix that respects the dimension of data
-* Fixed bug on condition for NA dispersion for models with 0 df resid
+* brmultinom returns a fitted values matrix that respects the
+  dimension of data.
+* Fixed bug on condition for NA dispersion for models with 0 df resid.
 
-# brglm 0.1.7
+# brglm2 0.1.7
 
 ## Other improvements, updates and additions
-* Eliminated errors from markdown chunks in multinomial vignette
+* Eliminated errors from markdown chunks in multinomial vignette.
 
 # brglm2 0.1.6
 
 ## Bug fixes
-* Compatibility with new version of enrichwith
+* Compatibility with new version of enrichwith.
 
 ## Other improvements, updates and additions
-* New email for Ioannis Kosmidis
+* New email for Ioannis Kosmidis.
 
 # brglm2 0.1.5
 
 ## Bug fixes
 
 ## New functionality
-* Added `type = AS_mixed` as an option to use **mean-bias reducing score functions** for the regression parameters and **median-bias reducing score functions** for the dispersion in models with unknown dispersion
-* `check_infinite_estimates` now accepts `brmultinom` objects
-* Added `singular.ok` argument to `brglmFit` and `detect_separation` methods in line with the update of `glm.fit`
+* Added `type = AS_mixed` as an option to use **mean-bias reducing
+  score functions** for the regression parameters and **median-bias
+  reducing score functions** for the dispersion in models with unknown
+  dispersion.
+* `check_infinite_estimates` now accepts `brmultinom` objects.
+* Added `singular.ok` argument to `brglmFit` and `detect_separation`
+  methods in line with the update of `glm.fit`.
 
 ## Other improvements, updates and additions
-* less strict tolerance in `brglm_control`
-* Updates to help files
-* Fixed typos in iteration vignette
-* Added URL and bugreports in Description
-* Added new tests
+* less strict tolerance in `brglm_control`.
+* Updates to help files.
+* Fixed typos in iteration vignette.
+* Added URL and bugreports in Description.
+* Added new tests.
 
 # brglm2 0.1.4
 
 ## Bug fixes
-* `brglmControl` is now exported
-* `slowit` did nothing; now included in iteration
+* `brglmControl` is now exported.
+* `slowit` did nothing; now included in iteration.
 
 ## New functionality
-* The `detect_separation` `method` for the `glm` function can be used to check for separation in binomial response settings without fitting the model. This relies on a port of Kjell Konis' `safeBinaryRegression:::separator` function (see ?detect_separation)
-* brglm2 provides estimation via **median-bias reducing score functions** with `type = "AS_median"`
-* brglm2 provides camel and underscored aliases for basic methods (`brglmFit`, `brglm_fit`, `detectSeparation`, `detect_separation`, `brglm_control`, `brglmControl`, `detectSeparationControl`, `detect_separation_control`, `checkInfiniteEstimates`, `check_infinite_estimates`)
+* The `detect_separation` `method` for the `glm` function can be used
+  to check for separation in binomial response settings without
+  fitting the model. This relies on a port of Kjell Konis'
+  `safeBinaryRegression:::separator` function (see ?detect_separation).
+* **brglm2** provides estimation via **median-bias reducing score
+  functions** with `type = "AS_median"`.
+* **brglm2** provides camel and underscored aliases for basic methods
+  (`brglmFit`, `brglm_fit`, `detectSeparation`, `detect_separation`,
+  `brglm_control`, `brglmControl`, `detectSeparationControl`,
+  `detect_separation_control`, `checkInfiniteEstimates`,
+  `check_infinite_estimates`).
 
 ## Other improvements, updates and additions
-* Minor enhancements in the codebase
-* The inverse expected information matrix is computed internally using cho2inv
-* Internal changes to have more meaningful variable names
-* Renamed detect_infinite* to check_infinite
+* Minor enhancements in the codebase.
+* The inverse expected information matrix is computed internally using
+  `cho2inv`.
+* Internal changes to have more meaningful variable names.
+* Renamed detect_infinite* to check_infinite.
 
 # brglm2 0.1.3
 
@@ -117,11 +162,11 @@
 
 ## Other improvements, updates and additions
 * Fixed typo in f_{Y_i}(y) in iteration vignette (thanks to Eugene
-  Clovis Kenne Pagui for spotting)
+  Clovis Kenne Pagui for spotting),
 
 # brglm2 0.1.2
 
-* First release
+* First release.
 
 
 

R/brglm2-package.R

@@ -94,7 +94,24 @@ if (getRversion() >= "2.15.1") globalVariables(c("n", "lambda"))
 #' @param object a fitted model object (e.g. the result of a
 #'     \code{\link{glm}} call).
 #' @param ... other options to be passed to the method.
+#'
+#'
+#' @note
+#'
+#' 
+#' \code{check_infinite_estimates} will be removed from \pkg{brglm2}
+#' at version 0.8. An new version of
+#' \code{check_infinite_estimates} is now maintained in the
+#' \pkg{detectseparation} R package at
+#' \url{https://cran.r-project.org/package=detectseparation}.
+#'
+#' @seealso check_infinite_estimates.glm
+#' 
 #' @export
 check_infinite_estimates <- function(object, ...) {
+    function_moves_to_new_package(gsub("\\(|\\)", "", deparse(match.call()[1])),
+                                  "0.8",
+                                  "brglm2",
+                                  "detectseparation")
     UseMethod("check_infinite_estimates")
 }

R/brglmControl.R

@@ -20,7 +20,6 @@
 #' Typically only used internally by \code{\link{brglmFit}}, but may
 #' be used to construct a \code{control} argument.
 #'
-#' @inheritParams stats::glm.control
 #' @aliases brglm_control
 #' @param epsilon positive convergence tolerance epsilon. Default is
 #'     \code{1e-06}.

R/brglmFit.R

@@ -1254,16 +1254,24 @@ confint.brglmFit <- function(object, parm, level = 0.95, ...) {
 #' Return the variance-covariance matrix for the regression parameters
 #' in a \code{\link{brglmFit}} object
 #'
-#' @inheritParams stats::vcov
+#' @inheritParams stats::vcov.glm
 #' @param model character specifying for which component of the model coefficients should be extracted.
 #'
+#' @details
+#'
+#' The options for \code{model} are \code{"mean"} for mean regression
+#' parameters only (default), \code{"dispersion"} for the dispersion
+#' parameter (or the transformed dispersion; see
+#' \code{\link{brglm_control}}), and \code{"mean"} for both the mean
+#' regression and the (transformed) dispersion parameters.
+#'
 #' @method vcov brglmFit
 #' @export
-vcov.brglmFit <- function(object, model = c("mean", "full", "dispersion"), ...) {
+vcov.brglmFit <- function(object, model = c("mean", "full", "dispersion"), complete = TRUE, ...) {
     model <- match.arg(model)
     switch(model,
            mean = {
-               summary.brglmFit(object, ...)$cov.scaled
+               vcov(summary.brglmFit(object, ...), complete = complete)
            },
            dispersion = {
                vtd <- 1/object$info_transformed_dispersion
@@ -1272,7 +1280,7 @@ vcov.brglmFit <- function(object, model = c("mean", "full", "dispersion"), ...)
                vtd
            },
            full = {
-               vbetas <- summary.brglmFit(object, ...)$cov.scaled
+               vbetas <- vcov(summary.brglmFit(object, ...), complete = complete)
                vtd <- 1/object$info_transformed_dispersion
                nBetasAll <- c(rownames(vbetas), paste0(object$transformation, "(dispersion)"))
                vBetasAll <- cbind(rbind(vbetas, 0),

R/brmultinom.R

@@ -46,7 +46,7 @@
 #' baseline-category logit models have infinite components with
 #' positive probability, which can result in problems in their
 #' estimation and the use of inferential procedures (e.g. Wad
-#' tests). Albert and Andreson (1984) have categorised the possible
+#' tests). Albert and Andreson (1984) have categorized the possible
 #' data patterns for such models into the exclusive and exhaustive
 #' categories of complete separation, quasi-complete separation and
 #' overlap, and showed that infinite maximum likelihood estimates
@@ -264,7 +264,7 @@ fitted.brmultinom <- function(object, ...) {
 #' @details
 #'
 #' The residuals computed are the residuals from the equivalent
-#' Poisson log-linear model fit, organised in a form that matches the
+#' Poisson log-linear model fit, organized in a form that matches the
 #' output of \code{fitted(object, type = "probs")}. As a result, the
 #' output is residuals defined in terms of the object and expected
 #' multinomial counts.
@@ -418,7 +418,7 @@ print.summary.brmultinom <- function(x, digits = x$digits, ...)
 #'     \code{"class"}, which produces predictions of the response
 #'     category at the covariate values supplied in \code{"newdata"},
 #'     selecting the category with the largest probability; the
-#'     alternative \code{"probs"} returns all cateogry probabilities
+#'     alternative \code{"probs"} returns all category probabilities
 #'     at the covariate values supplied in \code{"newdata"}.
 #' @param ... further arguments passed to or from other methods.
 #'

R/check_infinite_estimates.R

@@ -28,7 +28,7 @@
 #'
 #' @details
 #'
-#' \code{check_infinite_estimates} attempts to identify the occurence
+#' \code{check_infinite_estimates} attempts to identify the occurrence
 #' of infinite estimates in GLMs with binomial responses by
 #' successively refitting the model. At each iteration the maximum
 #' number of allowed IWLS iterations is fixed starting from 1 to
@@ -43,6 +43,16 @@
 #' estimate for the corresponding parameter has value minus or plus
 #' infinity.
 #'
+#' @note
+#' 
+#' \code{check_infinite_estimates} will be removed from \pkg{brglm2}
+#' at version 0.8. An new version of \code{check_infinite_estimates}
+#' is now maintained in the \pkg{detectseparation} R package at
+#' \url{https://cran.r-project.org/package=detectseparation}. In order
+#' to use the version in \code{detect_separation} load first
+#' \pkg{brglm2} and then \pkg{detectseparation}, i.e.
+#' \code{library(brglm2); library(detectseparation)}.
+#'
 #' @seealso \code{\link[nnet]{multinom}}, \code{\link{brmultinom}}
 #'
 #' @references
@@ -73,7 +83,7 @@
 #' }
 #' @export
 check_infinite_estimates.glm <- function(object, nsteps = 20, ...)
-{
+{   
     is_brmultinom <- inherits(object, "brmultinom")
 
     if ((class(object)[1] != "glm") & (!is_brmultinom)) {