Package {IncrementalityTEST}

Title:	Analyze Incrementality Experiments
Version:	0.1.1
Description:	Tools for calculating commerce metrics, pairing treatment and control results from A/B testing experiments, estimating incremental effects, and quantifying uncertainty with Student's t and nonparametric bootstrap confidence intervals. Includes validation helpers, a high-level analysis workflow, and compatibility functions for the original package interface.
License:	Apache License (≥ 2)
URL:	https://github.com/vkobayashi/IncrementalityTEST
BugReports:	https://github.com/vkobayashi/IncrementalityTEST/issues
Depends:	R (≥ 4.1.0)
Suggests:	boot (≥ 1.3-18), knitr, rmarkdown, testthat (≥ 3.0.0)
VignetteBuilder:	knitr
Config/testthat/edition:	3
Encoding:	UTF-8
Config/roxygen2/version:	8.0.0
NeedsCompilation:	no
Packaged:	2026-07-30 08:01:36 UTC; vladimerkobayashi
Author:	Vladimer Kobayashi [aut, cre]
Maintainer:	Vladimer Kobayashi <vladimer.kobayashi@gmail.com>
Repository:	CRAN
Date/Publication:	2026-08-07 16:30:25 UTC

Analyze an incrementality experiment collection

Description

A high-level workflow that pairs groups, calculates experiment-level differences, and summarizes the overall effect with t and bootstrap confidence intervals.

Usage

analyze_incrementality(
  data,
  metric,
  id_col = "experiment",
  group_col = "group",
  control = "control",
  treatment = "treatment",
  direction = c("treatment-control", "control-treatment"),
  na_action = c("error", "omit"),
  conf_level = 0.95,
  bootstrap_times = 2000L,
  seed = NULL
)

Arguments

Value

An object of class incrementality_analysis.

Examples

results <- data.frame(
  experiment = rep(LETTERS[1:4], each = 2),
  group = rep(c("control", "treatment"), 4),
  revenue_per_user = c(10, 11, 8, 8.8, 12, 13, 9, 10.5)
)
analyze_incrementality(
  results, "revenue_per_user",
  bootstrap_times = 199, seed = 42
)

Bootstrap confidence interval for a mean effect

Description

Uses ordinary nonparametric resampling of experiment-level differences.

Usage

bootstrap_incrementality(
  x,
  times = 2000L,
  conf_level = 0.95,
  type = "percentile",
  seed = NULL,
  na_rm = TRUE
)

Arguments

Value

An object of class incrementality_bootstrap, represented as a list with the estimate, standard error, interval, and bootstrap replicates.

Examples

bootstrap_incrementality(c(0.4, 0.8, 0.1, 0.6), times = 199, seed = 1)

Calculate commerce metrics

Description

Calculates revenue per user (RPU), buyer rate (BR), average order value (AOV), and transactions per buyer (TPB). Inputs are recycled using base R rules, so the function works with individual totals or equally sized vectors.

Usage

calculate_metrics(
  nb_transactions,
  nb_users,
  revenue,
  nb_buyers,
  zero_denominator = c("na", "allow")
)

Arguments

Value

A data frame with columns RPU, BR, AOV, and TPB.

Examples

calculate_metrics(
  nb_transactions = 11278,
  nb_users = 297073,
  revenue = 279480.4,
  nb_buyers = 10909
)

Statistic for the bootstrap (legacy interface)

Description

Statistic for the bootstrap (legacy interface)

Usage

incrementality_func(datadiff, indices)

Arguments

Value

Mean and estimated variance of the mean.

Calculate commerce metrics (legacy interface)

Description

This compatibility wrapper returns the same named list as the original package API. New code should generally use calculate_metrics().

Usage

incrementality_metrics(nb_transactions, nb_users, revenue, nb_buyers)

Arguments

Value

A named list containing RPU, BR, AOV, and TPB.

Examples

incrementality_metrics(11278, 297073, 279480.4, 10909)

Legacy vector summary helpers

Description

These small helpers are retained for compatibility. They return NA when all values are missing. fnFreqDay() returns the first mode when tied.

Usage

fnFreqDay(x)

fnMax(x)

fnAve(x)

fnSum(x)

Arguments

Value

fnAve(), fnMax(), and fnSum() return a numeric scalar. fnFreqDay() returns the name of the most frequent value, as a character scalar.

Pair experiment groups and calculate metric differences

Description

For every experiment, pairs one control value with one treatment value and computes treatment - control by default. Each experiment must contain exactly one row for each requested group.

Usage

metric_differences(
  data,
  metric,
  id_col = "experiment",
  group_col = "group",
  control = "control",
  treatment = "treatment",
  direction = c("treatment-control", "control-treatment"),
  na_action = c("error", "omit")
)

Arguments

Value

A data frame with the experiment ID, control and treatment values, and a difference column.

Examples

results <- data.frame(
  experiment = rep(c("A", "B", "C"), each = 2),
  group = rep(c("control", "treatment"), 3),
  RPU = c(10, 11, 8, 9.5, 12, 11.5)
)
metric_differences(results, "RPU")

Run a bootstrap (legacy interface)

Description

Requires the suggested boot package and returns a boot object.

Usage

res_boot(mydatadiff, statisticfunc = incrementality_func, nb_boot = 2000L)

Arguments

Value

An object returned by boot::boot().

Student's t confidence interval for a mean effect

Description

Student's t confidence interval for a mean effect

Usage

t_confidence_interval(x, conf_level = 0.95, na_rm = TRUE)

Arguments

Value

A one-row data frame with the sample size, mean, standard error, confidence level, and lower and upper confidence limits.

Examples

t_confidence_interval(c(0.4, 0.8, 0.1, 0.6))

Student's t confidence interval (legacy interface)

Description

Student's t confidence interval (legacy interface)

Usage

t_test_cf(datawithID, confinter = 0.05)

Arguments

Value

A list containing the confidence interval, mean, and standard error.

Calculate metric differences (legacy interface)

Description

Compatibility wrapper for datasets containing iabtest_id and abt_group, where control is coded 0 and treatment is coded 1. It preserves the original convention of control minus treatment.

Usage

test_metric(mydata, metric)

Arguments

Value

A two-column data frame with iabtest_id and ⁠inc_<metric>⁠.