Skip to contents

Plot Phi Correlations Between Binary Variables

Source: R/PlotPhiHeatmap.R
PlotPhiHeatmap.Rd

Computes pairwise phi coefficients between binary categorical variables with explicit 0/1 coding (1 == PositiveLevel from createBinaryMapping()), then renders heatmap-style plots with raw and FDR-adjusted significance.

Usage

PlotPhiHeatmap(
  data,
  CatVars,
  Relabel = TRUE,
  binary_map = NULL,
  fdr_scope = c("matrix", "per_outcome", "per_predictor"),
  cluster_rows = FALSE,
  cluster_columns = FALSE,
  triangle = c("full", "upper", "lower"),
  IncludeTrendSymbol = FALSE,
  Data = lifecycle::deprecated()
)

Arguments

data

A dataframe.

CatVars

Character vector of binary categorical variable names.

Relabel

Logical; if TRUE, uses sjlabelled variable labels for axes.

binary_map

Optional mapping as returned by createBinaryMapping(). If NULL, a mapping is created internally for CatVars.

fdr_scope

Either "matrix" (default) or "per_outcome", passed to ApplyFDRCorrection(). "matrix" corrects across all p-values at once (historical behavior). "per_outcome" corrects separately within each y-axis variable (YVar); the Phi matrix is symmetric, so this treats each variable's row of tiles as one family.

cluster_rows, cluster_columns

Logical; cluster y-axis rows and/or x-axis columns using displayed Phi-coefficient profiles.

triangle

Display "full" (default), "upper", or "lower" half of the symmetric Phi matrix. This affects the plot only; returned results remain complete.

IncludeTrendSymbol

Logical; add a plus sign for 0.05 < p <= 0.10.

Data

Deprecated (since 19.15.0). Use data instead.

Value

A list with:

  • Unadjusted: list(PvalTable, plot)

  • FDRCorrected: list(PvalTable, plot)

  • method = "Phi"

  • Relabel

  • BinaryMapping (used)

Examples

data(SampleData)
data(SampleVariableTypes)

Labelled <- RevalueData(SampleData, SampleVariableTypes)$RevaluedData

# CatVars must be binary (exactly two unique non-NA values). Derive a few
# more binary indicators so the matrix has off-diagonal structure to read;
# self-associations on the diagonal are masked out.
Labelled$APOE4 <- ifelse(
  grepl("E4", as.character(Labelled$Genotype)), "Carrier", "Non-carrier")
Labelled$HighTau <- ifelse(
  Labelled$tau > stats::median(Labelled$tau, na.rm = TRUE), "High", "Low")
Labelled$LowAbeta <- ifelse(
  Labelled$Ab_42 < stats::median(Labelled$Ab_42, na.rm = TRUE), "Low", "High")

result <- PlotPhiHeatmap(
  Labelled,
  CatVars = c("Diagnosis", "sex", "APOE4", "HighTau", "LowAbeta")
)

# Raw p-value phi heatmap
result$Unadjusted$plot


# FDR-adjusted phi heatmap
result$FDRCorrected$plot


# Show each binary-variable pair once.
upper <- PlotPhiHeatmap(
  Labelled,
  CatVars = c("Diagnosis", "sex", "APOE4", "HighTau", "LowAbeta"),
  triangle = "upper"
)
upper$FDRCorrected$plot