More options in box_plot and chord_diagram_plot

R/box_plot.R

@@ -153,6 +153,7 @@ convert_names <- function(x,
 #' Only works when given a single fit.
 #' Should be a named vector/list, with the names corresponding to types, and the value equal to the class name.
 #' @param involving Optional vector/list of types. Only coefficients involving these types will be plotted.
+#' @param how If the `involving` argument is supplied, should it involve *only* those types, or at least *one* of those types (which is relevant if inter-type interactions are involved).
 #' @export
 #' @keywords internal
 #' @md
@@ -164,7 +165,8 @@ make_summary_df <- function(fits,
                             full_names,
                             compute_confidence_intervals,
                             classes,
-                            involving) {
+                            involving,
+                            how) {
   # Take care of the classes argument
   if(!is.null(classes)) {
     classes <- as.list(classes)
@@ -225,8 +227,15 @@ make_summary_df <- function(fits,
         types
       }
 
+      # Should other types be considered?
+      other_types_subset <- if(how == "only") {
+        types_subset
+      } else {
+        types
+      }
+
       if(identification != "beta") { # Create dataframe of two columns with all possible pairs of types
-        d <- as.data.frame(expand.grid(from = types, to = types_subset, stringsAsFactors = FALSE))
+        d <- as.data.frame(expand.grid(from = other_types_subset, to = types_subset, stringsAsFactors = FALSE))
         d <- d[!duplicated(t(apply(d, 1, sort))), ]
 
         if(which == "within") { # in this case, remove columns that are the same, so shows only between interactions
@@ -365,6 +374,7 @@ make_summary_df <- function(fits,
 #' Only works when given a single fit.
 #' Should be a named vector/list, with the names corresponding to types, and the value equal to the class name.
 #' @param involving Optional vector/list of types. Only coefficients involving these types will be plotted.
+#' @param how If the `involving` argument is supplied, should it involve *only* those types, or at least *one* of those types (which is relevant if inter-type interactions are involved).
 #' @param title Plot title.
 #' @param colours Optional vector of colours to represent the different fits/classes.
 #' @param highlight_zero Highlight the zero value with a red line?
@@ -403,13 +413,17 @@ box_plot <- function(...,
                      compute_confidence_intervals = TRUE,
                      classes = NULL,
                      involving = NULL,
+                     how = c("only", "one"),
                      title,
                      colours = c("black", "#5BBCD6", "#F2AD00", "#00A08A", "#FF0000"),
                      highlight_zero = TRUE,
                      text_size = 16,
                      base_size = 20,
                      xmin,
                      xmax) {
+  # Interpret the how argument
+  how <- match.arg(how)
+
   # TODO: Avoid NULL default arguments for args forwarded to internal function, just set those to NULL in there
   # TODO: Is there a clean way to copy the documentation between functions when they have the same arguments? See box_plot and chord_diagram, all the gibbsm rgibbs functions, etc.
   # Interpret the which argument
@@ -425,7 +439,8 @@ box_plot <- function(...,
                         full_names = full_names,
                         compute_confidence_intervals = compute_confidence_intervals,
                         classes = classes,
-                        involving = involving)
+                        involving = involving,
+                        how = how)
 
   # Remove NAs
   df <- df[rowSums(is.na(df)) == 0, ]

R/chord_diagram.R

@@ -16,6 +16,7 @@
 #' @param classes If this parameter is supplied, then colours are used to distinguish classes.
 #' Should be a named vector/list, with the names corresponding to types, and the value equal to the class name.
 #' @param involving Optional vector/list of types. Only coefficients involving these types will be plotted.
+#' @param how If the `involving` argument is supplied, should it involve *only* those types, or at least *one* of those types (which is relevant if inter-type interactions are involved).
 #' @param include_self Include self-interactions?
 #' @param show_legend Show legend(s)?
 #' @param cex Cex.
@@ -54,6 +55,7 @@ chord_diagram_plot <- function(fit,
                                compute_confidence_intervals = TRUE,
                                classes = NULL,
                                involving = NULL,
+                               how = c("only", "one"),
                                include_self = TRUE,
                                show_legend = TRUE,
                                cex = 1,
@@ -68,6 +70,9 @@ chord_diagram_plot <- function(fit,
                                big_gap = 5,
                                repulsion_attraction_colours = c("blue", "red"),
                                classes_colours) {
+  # Interpret the how argument
+  how <- match.arg(how)
+
   # Take care of repulsion_attraction_colours argument
   if(!is.character(repulsion_attraction_colours) | length(repulsion_attraction_colours) != 2) {
     stop("repulsion_attraction_colours should be a vector containing two strings, the first one a colour to represent negative interactions, the second one representing positive interactions.")
@@ -88,7 +93,8 @@ chord_diagram_plot <- function(fit,
                                    full_names = full_names,
                                    compute_confidence_intervals = compute_confidence_intervals,
                                    classes = classes,
-                                   involving = involving)
+                                   involving = involving,
+                                   how = how)
 
   # Remove NAs
   chord_diagram <- chord_diagram[rowSums(is.na(chord_diagram)) == 0, ]