Package index • evfuse

Package

evfuse evfuse-package: evfuse: Spatial Data Fusion for Extreme Value Analysis

Data

Load data and compute distance matrices

coast_data: U.S. coastal sea level data
prediction_grid: Coastal prediction grid
load_data(): Load and validate sea level data
compute_distances(): Compute pairwise great-circle distances between sites
compute_cross_distances(): Cross-distance matrix between two sets of sites

Stage 1: GEV Fitting

Pointwise generalized extreme value fits

fit_gev_all(): Fit GEV at all sites (Stage 1, stationary)
fit_gev_detrended(): Fit GEV at all sites with NOAA detrending (Stage 1, nonstationary)
fit_gev_ns(): Fit nonstationary GEV with covariate (Stage 1)
gev_return_level(): GEV quantile function (return level)
gev_exceedance_prob(): GEV exceedance probability

Bootstrap

Measurement uncertainty via block bootstrap

bootstrap_W(): Estimate W via nonparametric block bootstrap (stationary)
bootstrap_W_detrended(): Bootstrap W with NOAA detrending
bootstrap_W_ns(): Bootstrap W with nonstationary covariate
taper_W(): Apply covariance tapering to W
subset_W_bs(): Subset bootstrap covariance to a single data source
embed_W(): Embed W_tap into full parameter space

Covariance

Spatial covariance and tapering functions

exp_cov_matrix(): Exponential covariance matrix
wendland2(): Wendland 2 compactly supported covariance function
wendland_taper_matrix(): Wendland 2 taper correlation matrix
build_sigma(): Build the coregionalization covariance matrix
build_taper(): Build the taper matrix for W

Stage 2: Spatial Model

Joint coregionalization model fitting

fit_spatial_model(): Fit Stage 2 spatial model via MLE
fit_naive_model(): Fit a naive (single-source) 3-dim spatial model
verify_gradient(): Verify analytic gradient against numerical finite differences

Kriging & Prediction

Spatial prediction and return level estimation

predict_krig(): Predict at new locations via universal kriging
predict_krig_naive(): Predict at new locations from a naive model
compute_return_levels(): Compute return levels with confidence intervals
compute_return_levels_ns(): Compute nonstationary return levels with covariate

Validation

Leave-one-out cross-validation

loo_cv(): Leave-one-out cross-validation via Rasmussen & Williams (2006) shortcut
loo_summary(): Summarize LOO-CV results

Trend Diagnostics

Temporal trend testing

mann_kendall_test(): Mann-Kendall trend test with Sen's slope
gev_trend_test(): GEV trend test via likelihood ratio