A warning message is now printed if the MLR in the gating network (only once, for any EM iteration of any model) fails to converge in itmax[3] iterations. The default value of itmax[3] has been increased from 100L to 1000L. This is liable to slow down the internal calls to nnet::multinom, but it generally reduces the required number of EM iterations. Minor .version_above() and ifelse fixes. More minor vignette edits. g.itmax now passed to glm(). Minor speedup to MoE_mahala for univariate data. CRAN release.

Author: Keefe-Murphy

DESCRIPTION

@@ -1,8 +1,8 @@
 Package: MoEClust
 Type: Package
-Date: 2020-12-29
+Date: 2021-06-21
 Title: Gaussian Parsimonious Clustering Models with Covariates and a Noise Component
-Version: 1.3.3
+Version: 1.4.0
 Authors@R: c(person("Keefe", "Murphy", email = "keefe.murphy@mu.ie", role = c("aut", "cre"), comment = c(ORCID = "0000-0002-7709-3159")),
            person("Thomas Brendan", "Murphy", email = "brendan.murphy@ucd.ie", role = "ctb", comment = c(ORCID = "0000-0002-5668-7046")))
 Description: Clustering via parsimonious Gaussian Mixtures of Experts using the MoEClust models introduced by Murphy and Murphy (2020) <doi:10.1007/s11634-019-00373-8>. This package fits finite Gaussian mixture models with a formula interface for supplying gating and/or expert network covariates using a range of parsimonious covariance parameterisations from the GPCM family via the EM/CEM algorithm. Visualisation of the results of such models using generalised pairs plots and the inclusion of an additional noise component is also facilitated. A greedy forward stepwise search algorithm is provided for identifying the optimal model in terms of the number of components, the GPCM covariance parameterisation, and the subsets of gating/expert network covariates.
@@ -17,10 +17,10 @@ Imports:
     matrixStats (>= 0.53.1),
     mclust (>= 5.4),
     mvnfast,
-    nnet,
+    nnet (>= 7.3-0),
     vcd
 Suggests:
-    cluster,
+    cluster (>= 1.4.0),
     clustMD (>= 1.2.1),
     geometry (>= 0.4.0),
     knitr,

R/Functions.R

@@ -24,8 +24,8 @@
 #' \itemize{
 #' \item{Type: }{Package}
 #' \item{Package: }{MoEClust}
-#' \item{Version: }{1.3.3}
-#' \item{Date: }{2020-12-29 (this version), 2017-11-28 (original release)}
+#' \item{Version: }{1.4.0}
+#' \item{Date: }{2021-06-21 (this version), 2017-11-28 (original release)}
 #' \item{Licence: }{GPL (>=2)}
 #' }
 #'
@@ -42,9 +42,9 @@
 #' \donttest{data(ais)
 #' 
 #' # Fit two sets of models
-#' res1  <- MoE_clust(ais[,3:7], G=2, gating=~BMI, expert=~sex,
+#' res1  <- MoE_clust(ais[,3:7], G=2, gating= ~ BMI, expert= ~ sex,
 #'                    modelNames=c("VEE", "EVE", "VVE"), network.data=ais)
-#' res2  <- MoE_clust(ais[,3:7], G=2, equalPro=TRUE, expert=~sex,
+#' res2  <- MoE_clust(ais[,3:7], G=2, equalPro=TRUE, expert= ~ sex,
 #'                    modelNames=c("VEE", "EVE", "VVE"), network.data=ais) 
 #'         
 #' # Compare the best model from each set of results
@@ -89,7 +89,7 @@
   version <- read.dcf(path, "Version")
   name    <- read.dcf(path, "Package")
   if(interactive()) {
-    packageStartupMessage(paste("\n___  ___      _____ _____ _           _   \n|  \\/  |     |  ___/  __ \\ |         | |     Gaussian Parsimonious \n| .  . | ___ | |__ | /  \\/ |_   _ ___| |_   Clustering Models with\n| |\\/| |/ _ \\|  __|| |   | | | | / __| __|        Covariates and a\n| |  | | (_) | |___| \\__/\\ | |_| \\__ \\ |_          Noise Component\n\\_|  |_/\\___/\\____/ \\____/_|\\__,_|___/\\__|           version", version, "\n"))
+    packageStartupMessage(paste("\n___  ___      _____ _____ _           _   \n|  \\/  |     |  ___/  __ \\ |         | |     Gaussian Parsimonious \n| .  . | ___ | |__ | /  \\/ |_   _ ___| |_   Clustering Models with\n| |\\/| |/ _ \\|  __|| |   | | | | / __| __|\t  Covariates and a\n| |  | | (_) | |___| \\__/\\ | |_| \\__ \\ |_\t   Noise Component\n\\_|  |_/\\___/\\____/ \\____/_|\\__,_|___/\\__|\t     version", version, "\n"))
   } else   {
     packageStartupMessage("\nPackage ", sQuote(name), " version ", version, ".\n")
   }

R/MoEClust.R

@@ -161,7 +161,7 @@
 #' # Use barcode plots for the categorical/continuous pairs.
 #' # Magnify the size of scatter points assigned to the noise component.
 #' 
-#' resN  <- MoE_clust(ais[,3:7], G=2, gating=~SSF + Ht, expert=~sex,
+#' resN  <- MoE_clust(ais[,3:7], G=2, gating= ~ SSF + Ht, expert= ~ sex,
 #'                    network.data=ais, modelNames="EEE", tau0=0.1, noise.gate=FALSE)
 #'                    
 #' MoE_gpairs(resN, data.ind=c(1,2,5), cov.ind=c(3,1,2), 
@@ -195,7 +195,7 @@ MoE_gpairs.MoEClust <- function(res, response.type = c("points", "uncertainty",
   gate  <- setdiff(attr(net, "Gating"), both)
   expx  <- setdiff(attr(net, "Expert"), both)
   if(is.null(subset$show.map)) {
-    subset$show.map <- ifelse(length(unique(claSS))  == 1, FALSE, (G + !is.na(res$hypvol)) > 1)
+    subset$show.map <- (length(unique(claSS)) != 1) && ((G + !is.na(res$hypvol) > 1))
   } else {
     if(length(subset$show.map) > 1  ||
        !is.logical(subset$show.map))                  stop("'subset$show.map' should be a single logical indicator", call.=FALSE)
@@ -316,7 +316,7 @@ MoE_gpairs.MoEClust <- function(res, response.type = c("points", "uncertainty",
     } else                                            stop("'outer.margins' are not valid", call.=FALSE)
   }
   if(is.null(outer.labels)) {
-    lab1   <- switch(EXPR=names(x)[1L], MAP=1L, ifelse(res$d == 1, 1L, 2L))
+    lab1   <- switch(EXPR=names(x)[1L], MAP=1L, as.integer(res$d != 1) + 1L)
     lab2   <- switch(EXPR=names(x)[1L], MAP=2L, 1L)
     outer.labels$top    <- vector("logical", N)
     outer.labels$top[seq(lab1,  N, by=2)] <- TRUE
@@ -1529,7 +1529,7 @@ plot.MoEClust <- function(x, what=c("gpairs", "gating", "criterion", "loglik", "
       }
       width     <- box.width
       y.fix     <- y - origin
-      fix.y     <- ifelse(!identical(show.counts, FALSE), 80, 90)
+      fix.y     <- ifelse(!isFALSE(show.counts), 80, 90)
       y.fix     <- if(max(y.fix) > fix.y) (y.fix * fix.y)/max(y.fix) else y.fix
       if(reference) panel.abline(h=origin, col=reference.line$col, lty=reference.line$lty, lwd=reference.line$lwd, identifier=paste(identifier, "abline", sep="."))
       panel.rect(x=x, y=rep(origin, length(x)), col=col, border=border, lty=lty, lwd=lwd, width=rep(width, length(x)), height=y.fix, just=c("centre", "bottom"), identifier=identifier)

R/Plotting_Functions.R

@@ -5,6 +5,7 @@ __with Gating and Expert Network Covariates__
 __and a Noise Component__
 =======================================================
 
+## MoEClust v1.4.0 - (_12<sup>th</sup> release [minor update]: 2021-06-21_)
 ### New Features, Improvements, Big Fixes, & Miscellaneous Edits
 * Various edits to `MoE_stepwise()` (thanks, in part, to requests from Dr. Konstantinos Perrakis):  
   * Added `initialModel` arg. for specifying an initial model from which to begin the search,  
@@ -39,12 +40,18 @@ __and a Noise Component__
 * `expert_covar` gains the arg. `weighted` to ensure cluster membership probabilities are properly  
   accounted for in estimating the extra variability due to the component means: defaults to `TRUE`,  
   but `weighted=FALSE` is provided as an option for recovering the old (not recommended) behaviour.
+* A warning message is now printed if the MLR in the gating network ever fails to converge,  
+  prompting the user to modify the `itmax` arg. to `MoE_control`: the 3<sup>rd</sup> element of this arg. governs  
+  the maximum number of MLR iterations --- consequently, its default has been modified from `100` to  
+  `1000` (thanks to a prompt from Dr. Georgios Karagiannis), which has the effect of slowing down  
+  internal calls to `nnet::multinom` but generally reduces the required number of EM iterations. 
 * Minor fix to `MoE_compare` whenever the optimal model needs to be refitted.
 * Fixed conflict between `mclust::as.Mclust` & `MoEClust::as.Mclust`:  
 `as.Mclust.MoEClust` now works regardless of order in which `mclust` & `MoEClust` are loaded.
 * Stronger checks for variables in `gating` & `expert` formulas which are not found in `network.data`.
+* Minor speed-up to initialisation for univariate response data with expert network covariates.
 * Minor speed-ups to some other utility functions.
-* Minor documentation and vignette edits.
+* Minor documentation, vignette, and vignette styling edits.
 
 ## MoEClust v1.3.3 - (_11<sup>th</sup> release [patch update]: 2020-12-29_)
 ### New Features, Improvements, Bug Fixes, & Miscellaneous Edits
@@ -75,7 +82,7 @@ __and a Noise Component__
 * Package startup message now checks if newer version of package is available from CRAN.
 * Updated citation info after publication in _Advances in Data Analysis and Classification_.
 * Updated maintainer e-mail address.
-* Minor documentation, examples, and CRAN compliance + `mclust` compatability edits.
+* Minor documentation, examples, and CRAN compliance + `mclust` compatibility edits.
 
 ## MoEClust v1.3.1 - (_9<sup>th</sup> release [patch update]: 2020-05-12_)
 ### New Features, Improvements, Bug Fixes, & Miscellaneous Edits
@@ -94,7 +101,7 @@ __and a Noise Component__
     * `MoE_mahala` now correctly uses the covariance of `resids` rather than the response.
     * New `MoE_mahala` arg. `identity` allow use of Euclidean distance instead:  
     this argument can also be passed via `exp.init$identity` to `MoE_control`.
-    * Convergence of the initialisation procedure now explictly monitored & sped-up.
+    * Convergence of the initialisation procedure now explicitly monitored & sped-up.
     * Values of the criterion being minimised are now returned as an attribute.
     * The number of iterations of the initialisation algorithm are also returned as an attribute.
     * `MoE_control` arg. `exp.init$max.init` now defaults to `.Machine$integer.max`.
@@ -263,7 +270,7 @@ __and a Noise Component__
 * Final `linf` estimate now returned for optimal model when `stopping="aitken"` & G > 1.
 * Removed redundant extra M-step after convergence for models without expert covariates.
 * Removed redundant & erroneous `resid` & `residuals` args. to `as.Mclust` & `MoE_gpairs`.
-* `MoE_plotCrit`, `MoE_plotGate` & `MoE_plotLogLik` now invisibly return revelant quantities.
+* `MoE_plotCrit`, `MoE_plotGate` & `MoE_plotLogLik` now invisibly return relevant quantities.
 * Corrected degrees of freedom calculation for `G=0` models when `noise.init` is not supplied.
 * Fixed `drop_levels` to handle alphanumeric variable names and ordinal variables.
 * Fixed `MoE_compare` when a mix of models with and without a noise component are supplied.