Skip to contents

Recommend Test Statistics via Offline KB Matching

Source: R/advise_offline.R
recommend_stat.Rd

Evaluates the KB decision table (category "stat_choice" rules) against the diagnostic signals extracted from a fitted EventStudyTask or a precomputed es_diagnostics object, and returns a severity-ranked es_advice S3 object.

Usage

recommend_stat(x, provider = NULL, ...)

# Default S3 method
recommend_stat(x, provider = NULL, ...)

# S3 method for class 'EventStudyTask'
recommend_stat(x, provider = NULL, ...)

# S3 method for class 'es_diagnostics'
recommend_stat(x, provider = NULL, ...)

Arguments

x

A fitted EventStudyTask (after fit_model()) or a precomputed es_diagnostics object returned by es_diagnostics().

provider

Accepted but ignored in the offline path – present only so the Phase 7 call shape is forward-compatible. Default NULL.

...

Additional arguments (currently ignored).

Value

An S3 object of class "es_advice" – a named list with:

source

"offline_kb" (character).

is_deterministic

TRUE – advice is rule-based, not LLM-generated.

rules_matched

Named list of matched rule records (severity-ranked: "error" first, then "warning", then "info"), each with fields id, recommendation, citation (list of author/year/key/venue), severity, category.

diagnostics_ref

The es_diagnostics list that was evaluated (possibly computed on-the-fly from the task).

Details

No LLM provider, network connection, or API key is required. Both functions are the always-available offline grounding layer (ADV-08). The returned es_advice object has the same shape as the Phase 7 Advice contract, flagged is_deterministic = TRUE and source = "offline_kb".

See also

flag_robustness, es_diagnostics, es_kb

Other eventstudy-advisor: AnthropicProvider, CustomProvider, OpenAICompatProvider, ProviderBase, es_advise(), es_diagnostics(), es_kb(), flag_robustness(), provider()

Examples

# \donttest{
data(dieselgate)
task <- EventStudyTask$new(dieselgate$firm, dieselgate$index, dieselgate$request)
task <- run_event_study(task, ParameterSet$new())
advice <- recommend_stat(task)
print(advice)
#> Offline Event Study Advice
#> ==========================
#> Source:           offline_kb 
#> Deterministic:    TRUE 
#> Rules matched:    2 
#> 
#> [WARNING] KB-NONNORM-NONPAR  (citation: BrownWarner1985)
#>   Recommendation: Non-normality detected in estimation-window residuals for >= 50% of events (Shapiro-Wilk p < 0.05). Consider non-parametric alternatives: Sign Test or Rank Test (Corrado 1989) are robust to departures from normality. Brown & Warner (1985) document that parametric tests lose size control under non-normal return distributions. [Threshold: 50% of events — ASSUMED, adjustable] 
#> 
#> [WARNING] KB-VAR-INCREASE-BMP  (citation: BMP1991)
#>   Recommendation: High CAR dispersion (IQR > 0.10 or SD > 0.15) suggests event-induced variance increase, which inflates Patell Z rejection rates. Use the BMP (Boehmer-Musumeci-Poulsen) test, which standardizes by the event-window variance and is specifically designed for this case. [Thresholds: IQR > 0.10, SD > 0.15 — ASSUMED, adjustable] 
#> 
# }