Skip to contents

Plot correlations heatmap

Source: R/PlotCorrelationsHeatmap.R
PlotCorrelationsHeatmap.Rd

Computes correlations or partial correlations and plots a heatmap. Handles:

  • continuous + categorical covariates

  • labelled data

  • non-syntactic names

  • sparse real-world datasets

  • ordinal variables

  • partial correlations via residualization

Usage

PlotCorrelationsHeatmap(
  data,
  predictor_vars = NULL,
  outcome_vars = NULL,
  covariates = NULL,
  method = "pearson",
  Relabel = TRUE,
  Ordinal = lifecycle::deprecated(),
  TreatOrdinalAs = "Categorical",
  min_n = 3,
  eps = 1e-12,
  fdr_scope = c("matrix", "per_outcome", "per_predictor"),
  Data = lifecycle::deprecated(),
  xVars = lifecycle::deprecated(),
  yVars = lifecycle::deprecated(),
  covars = lifecycle::deprecated()
)

Arguments

data

data.frame

predictor_vars

character vector

outcome_vars

character vector

covariates

optional covariates

method

pearson/spearman/kendall

Relabel

use labels

Ordinal

Deprecated logical compatibility option; use TreatOrdinalAs instead.

TreatOrdinalAs

How ordinal variables are handled. "Continuous" includes ordinal scores and "Exclude" omits them.

min_n

minimum complete rows

eps

variance tolerance

fdr_scope

Either "matrix" (default) or "per_outcome", passed to ApplyFDRCorrection(). With "matrix", FDR correction is applied across the whole p-value matrix at once (historical behavior). With "per_outcome", correction is applied separately within each outcome: in this function outcomes are the columns of the p-value matrix, i.e. outcome_vars (outcome_margin = 2).

Data

Deprecated (since 19.15.0). Use data instead.

xVars

Deprecated (since 19.15.0). Use predictor_vars instead.

yVars

Deprecated (since 19.15.0). Use outcome_vars instead.

covars

Deprecated (since 19.15.0). Use covariates instead.

Value

A list. Unadjusted and FDRCorrected each contain r, p, npairs, and plot. The standardized aliases p (same as Unadjusted) and p_fdr (same as FDRCorrected) are also included.

Examples

data(SampleData)
data(SampleVariableTypes)

# Attach labels for readable axes
Labelled <- RevalueData(SampleData, SampleVariableTypes)$RevaluedData

# Square heatmap: the same 10 variables on both axes
vars <- c("age", "AXL", "Adiponectin", "Alpha_1_Antitrypsin",
          "Alpha_2_Macroglobulin", "Apolipoprotein_A1", "Apolipoprotein_B",
          "C_Reactive_Protein", "Cortisol", "Insulin")

square <- PlotCorrelationsHeatmap(
  Labelled,
  predictor_vars = vars,
  outcome_vars = vars
)
square$Unadjusted$plot

square$FDRCorrected$plot


# Rectangular heatmap: different variables on x and y
rectangular <- PlotCorrelationsHeatmap(
  Labelled,
  predictor_vars = c("age", "AXL", "Adiponectin", "Cortisol", "Insulin"),
  outcome_vars = c("Apolipoprotein_A1", "Apolipoprotein_B",
                   "C_Reactive_Protein", "Ferritin", "Leptin")
)
rectangular$Unadjusted$plot