Analyze Rate-Integration Relationships Across Regimes

Prepare the point sets and bootstrap confidence curves used for the Supplementary Figure 4A-style rate-vs-variance and rate-vs-integration panels. Use plot() on the returned object to draw the panels.

Usage

regime_integration_relationships(
  summaries,
  resid_sd_threshold_vars = 2,
  resid_sd_threshold_corrs = 2,
  n_boot = 1000,
  ci_level = 0.99,
  seed = NULL,
  min_tips = NULL
)

Arguments

summaries

A list of manuscript-compatible vars_cors tables or a data frame from summarize_regime_covariances(). Regime IDs within each input table must be unique and non-empty. List inputs are pooled with a run column; missing or blank list names become run1, run2, and so on based on their positions, and the normalized names must be unique.

resid_sd_threshold_vars, resid_sd_threshold_corrs

Non-negative finite studentized-residual thresholds used to filter the variance and correlation panels.

n_boot

Number of bootstrap replicates for confidence curves.

ci_level

Confidence level for bootstrap ribbons.

seed

Optional random seed for reproducible bootstrap curves.

min_tips

Optional minimum tip count for inclusion when summaries contain a tip_count column. Relationship summaries are retained when tip_count >= min_tips; this differs from the strict PCA manuscript filter used by regime_correlation_pca().

Value

An object of class regime_integration_relationships, containing plot-ready data frames, fitted lm objects, bootstrap curves, and settings.

Details

High-correlation filtering is applied before this step with summarize_regime_covariances() via remove_high_corr and corr_threshold. The relationship object records residual filters and bootstrap settings, but it does not reapply matrix-level correlation filters. Rows with incomplete data for a panel remain in combined with an NA studentized residual and are omitted from that panel's point and removed-row data frames. Non-missing rates and variances must be finite and strictly positive, and non-missing correlations must be finite values in [-1, 1]. Boundary correlations at -1 or 1 are retained in combined, but their undefined Fisher-Z transforms and correlation-panel residuals are NA.