EnTraineR (version 1.0.0)

trainer_AovSum: Trainer: Interpret ANOVA (AovSum) with an LLM-ready prompt

Description

Builds an English-only, audience-tailored prompt to interpret an ANOVA produced by FactoMineR::AovSum. The function never invents numbers: it only passes verbatim excerpts to the LLM and instructs how to interpret deviations (sum-to-zero coding) as performance drivers.

Usage

trainer_AovSum(
  aovsum_obj,
  introduction = NULL,
  alpha = 0.05,
  t_test = NULL,
  audience = c("beginner", "applied", "advanced"),
  summary_only = FALSE,
  llm_model = "llama3",
  generate = FALSE
)

Value

Character prompt (if generate = FALSE) or a list.

Arguments

aovsum_obj

An object whose printed output contains sections named "Ftest" and "Ttest" (e.g., FactoMineR::AovSum()).

introduction

Optional character context paragraph for the analysis. Defaults to a generic description.

alpha

Numeric significance level used as an instruction for the LLM. Default 0.05.

t_test

Optional character vector to filter the T-test section by factor names and/or interactions (e.g. "Factor A" or "Factor A:Factor B").

audience

Target audience, one of c("beginner","applied","advanced").

summary_only

Logical; if TRUE, return a compact 3-bullet executive summary.

llm_model

Character model name for the generator (e.g., "llama3").

generate

Logical; if TRUE, calls trainer_core_generate_or_return().

Examples

if (FALSE) {
# Example 1: SensoMineR chocolates (requires SensoMineR)
if (requireNamespace("SensoMineR", quietly = TRUE)) {
# Load data from SensoMineR
data("chocolates", package = "SensoMineR")
# ANOVA summary with Product and Panelist
res <- FactoMineR::AovSum(Granular ~ Product * Panelist, data = sensochoc)

intro <- "Six chocolates have been evaluated by a sensory panel,
during two days, according to a sensory attribute: granular.
The panel has been trained according to this attribute
and panellists should be reproducible when rating this attribute."
intro <- gsub("\n", " ", intro)
intro <- gsub("\\s+", " ", intro)
cat(intro)

prompt <- trainer_AovSum(res, audience = "beginner",
                 t_test = c("Product", "Panelist"),
                 introduction = intro)
cat(prompt)

res <- gemini_generate(prompt, compile_to = "html")
}

# Example 2: Poussin dataset (shipped with this package)
data(poussin)
intro <- "For incubation, 45 chicken eggs were randomly assigned to three batches of 15.
Three treatments (different incubation temperatures) were then applied to the batches.
We assume that after hatching, all chicks were raised under identical conditions
and then weighed at a standard reference age.
At that time, the sex of the chicks - a factor known beforehand to cause
significant weight differences - could also be observed.
The objective is to choose the treatment that maximizes chick weight."
intro <- gsub("\n", " ", intro)
intro <- gsub("\\s+", " ", intro)
cat(intro)

res <- FactoMineR::AovSum(Weight ~ Gender * Temperature, data = poussin)

prompt <- trainer_AovSum(res,
                 audience = "beginner",
                 t_test = c("Gender", "Temperature"),
                 introduction = intro)
cat(prompt)

res <- gemini_generate(prompt, compile_to = "html")
}