Package index • csalert

Ensemble format and naming grammar

The draw-parallel ensemble container, the reporting-triangle input format, and the self-describing column-naming grammar they share.

csfmt_ensemble_v3(): Construct a csfmt_ensemble_v3
validate_ensemble(): Check a csfmt_ensemble_v3's structural shape
print(<csfmt_ensemble_v3>): Print a `csfmt_ensemble_v3`
csfmt_reporting_triangle_v3(): Construct a csfmt_reporting_triangle_v3
reporting_triangle_matrix(): Densify a reporting triangle into per-series reference x delay count matrices
set_time_series_id(): Assign content-hash time_series_id (+ readable label) by reference
csfmt_var(): Construct a csfmt measure column name from components
csfmt_parse(): Parse a csfmt measure column name into components
csfmt_interpret(): Interpret a dataset's columns via the naming grammar
q_label(): Probability -> controlled-vocabulary quantile label
q_value(): Quantile label -> probability

Ensemble analysis pipeline

Stages that add draw columns to an ensemble, then collapse it to a quantile summary (optionally healed into a cstidy csfmt_rts_data_v3).

ens_add_rate(): Add a rate measure to an ensemble
ens_collapse(): Collapse a csfmt_ensemble_v3 to a quantile-summary
mem_thresholds_v1(): MEM intensity thresholds
rolling_slope_matrix(): Rolling OLS slope over a weeks x draws matrix

Nowcasting engines

Complete a right-truncated reporting triangle into a csfmt_ensemble_v3.

nowcast_quasipoisson_v1(): Nowcast a reporting triangle into an ensemble (quasipoisson reporting regression)
nowcast_passthrough_to_ensemble_v1(): Build an ensemble from a reporting triangle WITHOUT nowcasting (passthrough)

Nowcast diagnostics

Replay a method against what was known in the past and score it: interval coverage, point-estimate revision, and reporting completion.

nowcast_censor(): Censor a reporting triangle to what was known "as of" a past week
nowcast_truth(): The settled (eventually-observed) total per reference week
nowcast_backtest(): Replay a nowcast method across as-of weeks (backtest)
nowcast_evaluate_v1(): Evaluate nowcast method(s): interval coverage + point-estimate revision
reporting_completion_v1(): Empirical reporting-completion summary from a reporting triangle
reporting_completion_trend_v1(): Reporting-completion trend: the delay curve by year and recent months
compare_results(): Compare two collapsed csfmt result sets

Nowcast calibration

Measure how far an engine’s intervals are from nominal coverage, as a per-horizon scaling factor. A diagnostic to check an engine with; not applied to published numbers unless you choose to.

nowcast_estimate_calibration_v1(): Estimate a nowcast calibration from a backtest
nowcast_apply_calibration_v1(): Apply a nowcast calibration to quantile predictions
print(<nowcast_calibration>): Print a `nowcast_calibration`

Input quality control

Verdict-only checks on surveillance input and week-over-week revisions.

qc_surveillance_data_v1(): Quality-control checks on surveillance input data
qc_week_over_week_v1(): Week-over-week QC: settled-data integrity (A) + frontier status signal (B)

Signal detection

Detect anomalies and compute prediction thresholds from historical surveillance data.

signal_detection_hlm(): Detect signals using the historical limits method
prediction_interval(): Prediction thresholds
prediction_interval(<glm>): Prediction thresholds

Short-term trend estimation

Estimate the direction and magnitude of recent trends in a time series.

short_term_trend(): Determine the short term trend of a timeseries
short_term_trend_sts_v1(): Determine the short term trend of a surveillance time series

Data simulation

Generate synthetic baseline and outbreak data for testing and evaluation.

simulate_baseline_data(): Simulate baseline surveillance data
simulate_seasonal_outbreak_data(): Add seasonal outbreaks to simulated data
simulate_spike_outbreak_data(): Add spiked outbreaks to simulated data
add_holiday_effect(): Apply a public holiday effect to simulated data