From 16baa392ade30d5b81db6f025c9c5331c578f4f8 Mon Sep 17 00:00:00 2001
From: Jirayus <jirayus.jiar@gmail.com>
Date: Mon, 25 Mar 2019 18:46:05 +1100
Subject: [PATCH] Address issue #9 - Implement a function to remove constant
 metrics and categorical metrics - Implement a function to get correlation
 groups according to VarClus

---
 NAMESPACE                         |  1 +
 R/AutoSpearman.R                  | 27 ++++++++------------------
 R/check.constant.categorical.R    | 32 +++++++++++++++++++++++++++++++
 R/get.vc.R                        |  4 ++++
 R/get.vc.correlation.groups.R     | 30 +++++++++++++++++++++++++++++
 man/check.constant.categorical.Rd | 22 +++++++++++++++++++++
 man/get.vc.correlation.groups.Rd  | 22 +++++++++++++++++++++
 7 files changed, 119 insertions(+), 19 deletions(-)
 create mode 100644 R/check.constant.categorical.R
 create mode 100644 R/get.vc.correlation.groups.R
 create mode 100644 man/check.constant.categorical.Rd
 create mode 100644 man/get.vc.correlation.groups.Rd

diff --git a/NAMESPACE b/NAMESPACE
index 77799df..e3d370b 100644
--- a/NAMESPACE
+++ b/NAMESPACE
@@ -2,6 +2,7 @@
 
 S3method(plot,summarizedvc)
 export(AutoSpearman)
+export(check.constant.categorical)
 export(fit)
 export(get.importance)
 export(loadDefectDataset)
diff --git a/R/AutoSpearman.R b/R/AutoSpearman.R
index e4afac4..32ddc15 100644
--- a/R/AutoSpearman.R
+++ b/R/AutoSpearman.R
@@ -12,7 +12,7 @@
 #' @param vif.threshold a numeric for a threshold of VIF score (default = 5)
 #' @param verbose  TRUE for printing
 #' @keywords AutoSpearman
-#' @examples 
+#' @examples
 #' Data = loadDefectDataset('groovy-1_5_7','jira')
 #' AutoSpearman(dataset = Data$data, metrics = Data$indep)
 #' @export
@@ -22,25 +22,14 @@ AutoSpearman <-
            spearman.threshold = 0.7,
            vif.threshold = 5,
            verbose = F) {
+    # Check constant metrics and categorical metrics
+    metrics <- check.constant.categorical(dataset, metrics)
     
-    # Check constant metrics
-    constant <- apply(dataset[, metrics], 2, function(x) max(x) == min(x))
-    constant <- names(constant[constant == TRUE])
-    # Remove constant metrics
-    if(length(constant) > 0){
-      metrics <- metrics[!metrics %in% constant]
-    }
     
-    # Check categorical metrics
-    category <- sapply(dataset[, metrics], class)
-    category <- names(category[category=="character"])
-    # Remove categorical metrics from Spearman Analysis
-    if(length(category) > 0){
-      metrics <- metrics[!metrics %in% category]
-    }
+    spearman.metrics <-
+      get.automated.spearman(dataset, metrics, spearman.threshold, verbose)
+    AutoSpearman.metrics <-
+      stepwise.vif(dataset, spearman.metrics, vif.threshold, verbose)
     
-    spearman.metrics <- get.automated.spearman(dataset, metrics, spearman.threshold, verbose)
-    AutoSpearman.metrics <- stepwise.vif(dataset, spearman.metrics, vif.threshold, verbose)
-
-    return(c(AutoSpearman.metrics,category))
+    return(AutoSpearman.metrics)
   }
diff --git a/R/check.constant.categorical.R b/R/check.constant.categorical.R
new file mode 100644
index 0000000..ddca673
--- /dev/null
+++ b/R/check.constant.categorical.R
@@ -0,0 +1,32 @@
+#' Check for constant metrics and categorical metrics
+#' 
+#' @param dataset  a data frame for data
+#' @param metrics  a characters or a vector of characters for independent variables
+#' @keywords constant categorical
+#' @examples
+#' Data = loadDefectDataset('groovy-1_5_7','jira')
+#' check.constant.categorical(dataset = Data$data, metrics = Data$indep)
+#' @export
+check.constant.categorical <-
+  function(dataset,
+           metrics) {
+    # Check constant metrics
+    constant <-
+      apply(dataset[, metrics], 2, function(x)
+        max(x) == min(x))
+    constant <- names(constant[constant == TRUE])
+    # Remove constant metrics
+    if (length(constant) > 0) {
+      metrics <- metrics[!metrics %in% constant]
+    }
+    
+    # Check categorical metrics
+    category <- sapply(dataset[, metrics], class)
+    category <- names(category[category == "character"])
+    # Remove categorical metrics from Spearman Analysis
+    if (length(category) > 0) {
+      metrics <- metrics[!metrics %in% category]
+    }
+    
+    return(metrics)
+  }
\ No newline at end of file
diff --git a/R/get.vc.R b/R/get.vc.R
index 772e3a0..6a59838 100644
--- a/R/get.vc.R
+++ b/R/get.vc.R
@@ -9,6 +9,10 @@
 #' @keywords VarClus
 
 get.vc <- function(dataset, metrics, similarity = 'spearman', varclus.threshold = 0.7){
+  
+  # Check constant metrics and categorical metrics
+  metrics <- check.constant.categorical(dataset, metrics)
+  
   f <- as.formula(paste("~", paste(metrics, collapse = " + ")))
   vc <-
     Hmisc::varclus(f,
diff --git a/R/get.vc.correlation.groups.R b/R/get.vc.correlation.groups.R
new file mode 100644
index 0000000..14dc7fe
--- /dev/null
+++ b/R/get.vc.correlation.groups.R
@@ -0,0 +1,30 @@
+#' Get correlation groups according to VarClus based on the absolute Spearman correlation coefficients between metrics
+#'
+#' This function makes life simple by providing a VarClus.
+#' @param dataset  a data frame for data
+#' @param metrics  a vector of characters or a vector of characters for independent variables
+#' @param similarity a character for similarity measures (e.g., Spearman rank correlation), default = spearman
+#' @param varclus.threshold a numeric for correlation coefficient threshold value
+#' @importFrom Hmisc varclus
+#' @keywords VarClus
+
+get.vc.correlation.groups <- function(dataset, metrics, similarity = 'spearman', varclus.threshold = 0.7){
+  
+  # Check constant metrics and categorical metrics
+  metrics <- check.constant.categorical(dataset, metrics)
+  
+  f <- as.formula(paste("~", paste(metrics, collapse = " + ")))
+  vc <-
+    Hmisc::varclus(f,
+                   similarity = similarity,
+                   data = dataset[, metrics],
+                   trans = "abs")
+  
+  var.clusters <-
+    cutree(vc$hclust, h = (1 - varclus.threshold))
+  melted.data <- melt(var.clusters)
+  varclus.correlation.groups <- data.frame(metrics = row.names(melted.data), rank = var.clusters)
+  row.names(varclus.correlation.groups) <- NULL
+  
+  return(varclus.correlation.groups)
+}
\ No newline at end of file
diff --git a/man/check.constant.categorical.Rd b/man/check.constant.categorical.Rd
new file mode 100644
index 0000000..b4c45cf
--- /dev/null
+++ b/man/check.constant.categorical.Rd
@@ -0,0 +1,22 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/check.constant.categorical.R
+\name{check.constant.categorical}
+\alias{check.constant.categorical}
+\title{Check for constant metrics and categorical metrics}
+\usage{
+check.constant.categorical(dataset, metrics)
+}
+\arguments{
+\item{dataset}{a data frame for data}
+
+\item{metrics}{a characters or a vector of characters for independent variables}
+}
+\description{
+Check for constant metrics and categorical metrics
+}
+\examples{
+Data = loadDefectDataset('groovy-1_5_7','jira')
+check.constant.categorical(dataset = Data$data, metrics = Data$indep)
+}
+\keyword{categorical}
+\keyword{constant}
diff --git a/man/get.vc.correlation.groups.Rd b/man/get.vc.correlation.groups.Rd
new file mode 100644
index 0000000..c729938
--- /dev/null
+++ b/man/get.vc.correlation.groups.Rd
@@ -0,0 +1,22 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/get.vc.correlation.groups.R
+\name{get.vc.correlation.groups}
+\alias{get.vc.correlation.groups}
+\title{Get correlation groups according to VarClus based on the absolute Spearman correlation coefficients between metrics}
+\usage{
+get.vc.correlation.groups(dataset, metrics, similarity = "spearman",
+  varclus.threshold = 0.7)
+}
+\arguments{
+\item{dataset}{a data frame for data}
+
+\item{metrics}{a vector of characters or a vector of characters for independent variables}
+
+\item{similarity}{a character for similarity measures (e.g., Spearman rank correlation), default = spearman}
+
+\item{varclus.threshold}{a numeric for correlation coefficient threshold value}
+}
+\description{
+This function makes life simple by providing a VarClus.
+}
+\keyword{VarClus}