Visualize multi-orientation environmental-species correlation heatmaps2

Usage

gglink_heatmap_triple(
  Environment,
  Experiment,
  edge,
  node,
  sample_col = "Sample",
  delim = ",",
  hub_n = NULL,
  r = 6,
  cor.method = c("pearson", "kendall", "spearman"),
  cor.use = c("pairwise", "everything", "all", "complete", "na"),
  env_p_adjust = "none",
  link_p_adjust = "none",
  sig_breaks = c(0.05, 0.01, 0.001)
)

Arguments

Environment: character or data.frame File path or data frame of environment data.
Experiment: character or data.frame File path or data frame of experiment data.
edge: character or data.frame File path or data frame of edge data. Must contain columns from and to; an optional numeric weight column is mapped to edge colour/width (defaults to 1 when absent).
node: character or data.frame File path or data frame of node data. Must contain a node column listing every node referenced by edge; node names matching columns of Experiment become the hub nodes anchored on the central heatmap. An optional annotation column drives node fill/shape (when absent it is derived automatically: "Experiment" for hub nodes, "Environment" otherwise).
sample_col: Character (default = "Sample") Column name used as sample ID when input is a data frame or file.
delim: Character (default = ",") Delimiter for reading input files.
hub_n: Integer (default = NULL) If NULL (recommended), hubs are the Experiment variables present in node. If an integer, the hub_n highest out-degree nodes are used instead (they must then correspond one-to-one to the Experiment variables, and node rows must list circle nodes first).
r: numeric (default = 6) Radius of the outer node circle.
cor.method: Character (default = "pearson") Correlation method passed to psych::corr.test(), used for both the Environment x Environment heatmap and the Environment x Experiment links. One of "pearson", "kendall", "spearman".
cor.use: Character (default = "pairwise") Missing-value handling passed to psych::corr.test(); same vocabulary as gglink_heatmaps(). Note the default differs from that function ("everything") because psych::corr.test() itself defaults to "pairwise", which is what this plot has always used.
env_p_adjust: Character (default = "none") Multiple-testing correction for the Environment x Environment correlations (the significance stars on the triangular heatmap). Any method accepted by stats::p.adjust(), or "none".
link_p_adjust: Character (default = "none") Multiple-testing correction for the Environment x Experiment correlations (the linetype of the link segments). Any method accepted by stats::p.adjust(), or "none". Note that psych::corr.test() defaults to "holm" here but still reports raw p-values in $p, so the previous hard-coded call was in effect uncorrected; "none" keeps that behaviour.
sig_breaks: Numeric vector of length 3 (default = c(0.05, 0.01, 0.001)) Strictly decreasing p-value cut points shared by the heatmap stars ("" / "*" / "**" / "***") and the link-segment linetype legend.

Value

a ggplot2 object

Examples

if (FALSE) { # \dontrun{
# Environment / Experiment: samples in rows (with a Sample column),
# variables in columns. Edges connect Experiment variables (hubs) to
# any other nodes.
p <- gglink_heatmap_triple(
  Environment = env_df,   # Sample + environmental variables
  Experiment  = exp_df,   # Sample + experiment variables (become hubs)
  edge        = data.frame(from = c("ExpA", "ExpB"),
                           to   = c("pH", "TN"),
                           weight = c(0.8, 0.5)),
  node        = data.frame(node = c("pH", "TN", "ExpA", "ExpB"))
)
} # }