NEWS

v0.4.1

2026-08-19

Bug Fixes

chglog: parse scoped conventional commits and show the scope
ci: pin full history for git-chglog and fail on an empty NEWS section
cran: restore the pkgdown site URL now that the site is live
cran: drop the unpublished pkgdown site URL from DESCRIPTION
preview: never open the data viewer while R CMD check runs
release: abort when NEWS.md has no notes for the version being released

Features

replication: add harness for replicating IES theses with artma

v0.4.0

2026-08-19

Bug Fixes

keep lintr in Suggests, it is used by the linter unit tests
drop LazyData and dead Suggests, honour absolute graphics export path
bma: drop non-numeric moderators instead of failing inside scale()
bpe: gate the “run BMA first?” prompt on the autonomy level
data: derive study_label from a legible study name column when study keys are numeric
data: write mock data to a durable user dir instead of tempfile
exogeneity: default the IV instrument to 1/sqrt(n_obs) (#450)
linear: make between-effects standard errors robust and label every model’s vcov
methods: keep sequential method output on the console so prompts show
output: add a closing run summary and stop reading integers as percentages
output: resolve absolute graphics path in writers and escape literal braces in help
output: normalise manifest paths to forward slashes for Windows
output: drop zero counts from the run summary and winsorization notes
report: split method skip contract into skip_reason and skipped_models

Features

api: rename cli.* to cli_* and keep cli.run as a permanent alias
api: rename autonomy, viz, methods, report, results exports to snake_case
api: rename config.* and data.preview with deprecated aliases
api: rename options.* exports to options_* with deprecated aliases
api: add lifecycle dependency and snake_case API naming guide
core: add leveled verbosity log wrappers, fix drifted defaults
methods: show the registered method metadata in methods.list()
options: make options discoverable via help tree, list details and diff
output: write a per-run run.json and drive the report and CLI from it
output: roll the estimates slot out across the remaining runtime methods
output: add an unrounded tidy estimates slot to the method contract

v0.3.5

2026-08-09

Bug Fixes

replace the homemade digest hash and stop the data cache invalidating itself
align LCM defaults and wire chunk_size option in p_hacking_tests
let column config data_type override type detection (#253)
stop false study_id warning on valid string labels (#250)
declare rjags in Suggests for the RoBMA test guard
target the RoBMA 4.x API in the robma method
make data config keys deterministic and invalidate the config df cache
update the climenu version to 0.1.5 to avoid an error
handle NULL var_name in config during column validation
silence the bma and fma run if the verbose_output is FALSE
prime data-config cache when data is passed directly to artma()
quiet dependency and validation noise in real runs and logs
requirements on bma-dependent methods
strict bpe requirements
schema reconcile
keep MAIVE pin CRAN-resolvable and gate tests on the runtime floor
require MAIVE 0.2.5 for unrounded estimates
install MAIVE from CRAN instead of GitHub
schema reconciliation messages
make BPE recommendation fallback an explicit no-recommendation outcome
declare runtime dependencies correctly
locate linters.R without the here package
recompute user-supplied precision after winsorization
locate the package root when the checkout is not named artma
implement puniform_method P (method of moments) branch
correct FE intercept SE vcov and study-weighted OLS weights in linear tests
use inverse-variance WAAP pilot weights and clustered WLS standard errors
isolate runtime method failures and always export partial results
quoted data paths, duplicate column mapping, exogeneity table gaps
prevent bma/best_practice_estimate crash on small BMA models
failing tests
options correctness fixes and dead code removal
draw mock standard errors from a strictly positive range
honor the cache TTL option and add a template-to-code option parity test
replace literal unicode escape text with ascii equivalents
force sequential execution when already inside a forked worker
restore RoxygenNote 7.3.3 and man page dropped by roxygen 8 regeneration
drop unused fs Suggests dependency and its dir_create lint rule
bma: align VIF masks, split multi-model params, keep unmatched regressor names
bma: exclude derived effect/se encodings from auto-selected moderators
bma: skip gracefully instead of crashing with a single moderator
bpe: apply overrides and report estimates on the raw effect scale
bpe: auto-detect dummy predictors for factor summary table
cache: record base-graphics plots so cached reruns restore them
ci: quote E2E test-file array so both fixture tests actually run
data: abort on colliding column-rename targets in standardize_column_names
data: treat NA options as unset and drop non-numeric required NAs under remove
data: silence duplicate drop_conflicting_raw log in configure phase
data: stop strategy no longer aborts on optional-column NAs
data: only hard-require columns the requested methods need
data: guard imputation of mostly-missing columns and drop constant BMA moderators
data: import get_verbosity in standardize_column_names, update stale test calls
data: stop treating lowercase na/n-a strings as missing values
data: default zero-SE handling to remove-with-warning instead of abort
detection: detect dummy/categorical type even with group_category set
e2e: shell-quote CLI args so the Rscript subprocess starts under sh
effect-summary-stats: correct weighted-mean SE, per-study weights, unrounded CI
effect_summary_stats: default effect column when nothing configured
elliott: use integer knot indices in pure R CDF simulation fallback
endo-kink: use model df, not residual df, in heterogeneity variance divisor (#371)
exogeneity: use se^2 as p-uniform* study variance and fix selection threshold
exogeneity: fill formatted columns when p-uniform* skips estimation
exogeneity: stop p-uniform* LRT collapsing to L=0, fall back to P on ML failure
exogeneity: rank IV instruments by first-stage F, drop backwards votes
exogeneity: cluster IV diagnostics and AR test, signed p-uniform search, true bias test
fma: correct Mallows criterion, Buckland SE centering, and data scaling
formatting: stop wide summary tables splitting into two blocks
formatting: apply wide-table print fix to remaining call sites
linear: raise bootstrap replications and flag CI/star disagreement
maive: guardrail AR interval when constant N disables instrument
maive: align se option with MAIVE 0.2.4, report effective spec and AR guardrails
maive: resolve add_significance_marks via shared helper restoring scalar check
maive: flag weak first-stage instrument and document maive_first_stage
modules: newline-terminate captured cli messages, NA-safe parallel option read
nonlinear: skip degenerate corner-solution fits instead of tabulating them
nonlinear: allow general selection cutoffs, cluster SEs by study, report pub probabilities
options: flatten nested user_input on create, honor mock input, fix CONST scope
options: merge nested user_input safely and validate before overwrite
options: normalize .yaml suffix when loading options file
options: keep the unified column store when migrating legacy options
options: stop options rewrites from dropping column records and value shapes
output: plain ellipsis, no raw NA cells, label standardized BPE values
output: resolve auto output dir per options file and never persist it
p-hacking: orient caliper tail, cluster by study, show direction
p-hacking: rewrite Cox-Shi against canonical Elliott et al. implementation
p_hacking: respecify caliper test as Gerber-Malhotra binomial test
p_hacking_tests: seed LCM/MAIVE, real cpp fallback, skip reasons
p_hacking_tests: quick wins from p-hacking suite review (#257)
p_hacking_tests: keep install/upgrade hints in MAIVE skip reason
preview: skip utils::View() when no data viewer is available
rng: strip BMS wall-clock reseed and key method RNG streams by name
run: cap forked workers at the environment’s process limit
run: pin BLAS/OpenMP to one thread while forking method layers
run: use a fork-safe graphics device so parallel runs can export plots
stem,robma: cluster RoBMA by study, align stem MSE table and plot rows
tests: probe an actual JAGS model fit to catch macOS oldrel crash
tests: avoid rjags segfault in robma JAGS probe on CI
tests: probe rjags for JAGS in a subprocess to avoid CI crash
variable_summary_stats: obs column counts non-missing observations
vcov: warn on silent clustering downgrades; sign-correct zero-SE t-stats
vcov: warn on no-cluster single-step downgrade to stats::vcov()
viz: skip plot previews in forked workers to avoid quartz fork crash
viz: probe the cairo device instead of trusting capabilities()

Code Refactoring

add shared method registration helper and standard return contract
make options.load pure and parse the options template once
box plot max boxes per plot option
simplify non-idiomatic constructs and reuse significance_mark
unify bma and fma into a single table
handle data columns through reconciliation
replace the cli replay cache layer with plain memoise
unify file-read dispatch and post-read normalization
collapse the 5-level autonomy system into a 3-value enum
vectorize row-wise loops and hoist repeated computation
unify required, reserved, and computed column name lists
declarative method metadata replacing execution order and dependency plumbing
unify weak-instrument F threshold constant in exogeneity.R
strict option access and fail-loud coercion
split prepare_data into configure, compute, and persist phases
consolidate R CMD check workarounds into one generated manifest
unify column mapping and data config into one per-column store
extract shared helpers for duplicated logic
remove dead functions, unused imports and unreachable branches
restore stats imports used via formula environments, with comment
bma: extract shared unwrap_bma_result from bpe and fma
bpe: split best_practice_estimate into econometric core and viz plots
data: drop dead interactive branch, move column mapping persistence
data: split schema_reconcile into detect, ui, persist layers
data: extract shared column profiler into data/profile.R
data: move na/se-zero configure resolvers into data/configure.R
econometric: route four vcov ladders through shared robust_vcov
elliott: make C++ CDF simulation the sole production path
formatting: consolidate nonlinear formatting helpers into shared modules
formatting: share one summary-table builder across econometric tests
methods: extract shared resolve_variable_groups equal/gltl resolver
options: add spec-driven option resolver and migrate method files
options: template-sourced typed option accessors
options: share one option type registry across coerce and validate
visualization: unify plot export and tick-interval math
visualization: drop ggtext, use plain element_text for ticks

Features

add opt-in RoBMA runtime method
add bpe
add BMA prior-comparison plot for multiple parameter sets
attach STEM funnel and MSE plots to nonlinear_tests output
support user-defined data subset conditions
add economic significance table to best_practice_estimate
add BPE per-factor density and sorted scatter plots
add BPE summary stats grouped by factor levels
bma: announce auto-included priority moderators during variable selection
cache: harden cache signature and enable caching by default
cli: add scriptable command-line interface with cli.run and cli.install
coverage: measure inst/artma box modules, upload merged report to codecov
data: reconcile column-mapping conflicts and give actionable collision errors
fma: add cluster-robust standard errors clustered by study_id
funnel_plot: report filtered observation count in subtitle and output
hooks: add local git hooks for style/lint and commit-msg checks
linear: precheck panel specs and skip with plain-language reasons
linear: enable bootstrap CIs for the between effects spec
maive: add automatic first-stage selection from sample-size spread
maive: split MAIVE into its own method with sectioned, annotated output
nonlinear: report effect heterogeneity (tau) for STEM, Hierarchical, and Endogenous Kink
output: add LaTeX table export via output.table_formats
p-hacking: make Elliott supports configurable and surface Cox-Shi skip reasons (#369)
p_hacking: raise default LCM iterations to 10000
report: add self-contained HTML report renderer via report.render()
robma: enable parallel MCMC chains when cores allow
run: run independent runtime methods in parallel layers
visualization: add optional interactive HTML plot export

Performance Improvements

bpe: split study row indices once instead of rescanning per study
elliott: share pair lambdas across Cox-Shi bound orders
funnel_plot: split rows by study once in aggregate_study_medians
linear: fast-path random effects bootstrap via Swamy-Arora replication
linear: fast-path bootstrap estimates without per-replication plm/vcov
linear: skip vcov math and trim columns in bootstrap replications
linear: hoist cluster split out of bootstrap loop
methods: cache only expensive stages via cached opt-in on register_runtime_method
p_hacking_tests: cache LCM critical-value simulation under a narrow key
prima_facie_graphs: vectorize group column and median split construction
stem: replace O(n^2) outer-product submatrix sums with cumsum
visualization: use ragg for faster PNG export when available

v0.3.3

2026-02-11

Bug Fixes

parallel cdfs workers function in the elliott model calculations
add missing global variables
description versino control
failing funnel plot tests
duplicate data reading during the main pipeline
attempt to remove extra printed rows
recover stale options files with template defaults
box imports around the elliott cpp functions
maive output formatting issue
set the artma open results default to false
remaining lints in the package
lints around the project
lint setup
elliott tests
elliott options
results.open folder path
fma option template duplicates
r cmd check notes/warnings
box_plot detection
failing tests
add proper visualization text to gg plots
return the main results of the artma method silently
mice algorithm na handling
pressing enter upon options value ask leads to file/folder choice
box plot rendering, autonomy level detection
update the methods return objects to print the output into console cleanly
test example lib imports
lints
remove failing tests
missing success print; create an options file if none exist when necessary
the interactive columns mapping now works
interactive mapping column selection
data reading during data preprocessing
more lint issues
indentation issues
ensure the system editors are available for editor opening
options edge cases
add computed columns to data config and suppress misleading warnings
use full path when saving preferences to options file
automatically remove aliased variables during BMA analysis
display variable reason correctly in suggestion output
correct data config import and add integration tests
remove automatic addition of ‘se’ to BMA variables
only exclude effect variable from BMA selection
update bma coefficient extraction
update the way bma objects are validated upon input
update the maive package reference
move AER and ivmodel from Imports to Suggests
handle vector values in option NA validation

Code Refactoring

make study ID column detection smarter
add a parallel cdfs computation
make the data config option sparse
unify and clean up visualization logic
unify study_id vs. study usage
clean up the autonomy system
make the follow-up user choice prompts respect autonomy
clean up the package structure
redirect user to r universe links in more help
unify the invocation public api methods to ‘artma’
the data source is now read in two-steps to enable column mapping
add better whitespace na handling when reading data
the na handling method now distinguishes between numeric and non-numeric columns
updated column position verification
the data preprocessing now removes empty columns based on quality, not position
simplify editor detection
change the maive installation to use the CRAN version
the bma variables selection is now saved to the data config node
modify bma variable selection; add prompts
add explanatory messages and fully silence clubSandwich warning
improve progress messages and silence clubSandwich warning
add Caliper tests to p-hacking method (partial integration)
remove 5 unnecessary dependencies with custom polyfills

Features

automatically open latest results through results.open by default
add prima-facie graphs
add rcpp, elliott cdfs faster calculation
persist the temporary results directories in the user option files
add a unified results handling
add FMA
add a t-stat histogram runtime method
add viz public methods
add a data.preview function
add funnel plot
add visualization module; box-plot
add help suppression for some option prompts
add the autonomy system
add a welcome message
implement the select all option of the climenu
add a public method options.delete
add options vignette, improve the ask/utils functions
allow the user to use a mock source DF
add auto columns detection
add better column type detection
add functionality for choosing a suitable editor
enhance optional column computation with comprehensive auto-calculation
add reusable save preference prompt for configuration choices
add comprehensive missing value handling system with smart prompts
add collinearity detection and priority mechanism for BMA variable suggestions
implement automatic variable suggestion for BMA
add interactive variable selection for effect summary stats
add make stats command for repository statistics
implement automatic variable suggestion for effect summary stats
implement interactive BMA variable selection
implement Bayesian Model Averaging (BMA) method
add winsorization to data preprocessing
add caliper_display_ratios option for Caliper tests
enhance p-hacking tests output formatting
integrate MAIVE package from GitHub
integrate Caliper, Elliott, and MAIVE as unified p-hacking tests
add p-hacking tests as a new runtime method

Performance Improvements

add progress bars and fix hash overflow warnings

v0.3.2

2025-10-07

Bug Fixes

disable auto-detection in standardize_column_names tests
reduce false positives in column recognition and fix options flow

Features

add exogeneity tests as a new runtime method
add a smart recognition system for input column names
add intelligent data reading and column recognition

v0.3.1

2025-10-06

Bug Fixes

rewrite the menu select usage to reference the climenu package correctly
udpate the fix_options.R script
resolve linter warnings and style issues

v0.3.0

2025-10-06

Bug Fixes

add Makefile and related files to .Rbuildignore
improve keyboard input handling in climenu

Code Refactoring

replace utils::select.list with climenu::menu

Features

replace run.sh with Makefile for build tasks
add climenu sub-package for interactive CLI menus

v0.2.2

2025-10-01

Bug Fixes

satisfy R CMD check warnings
the p-value significance mark formatting is now more robust
tidy linear tests output table formatting
align linear tests runtime behaviour
update a number of custom methods and their tests
isolate cache keys per runtime stage
update broken code in the effect and summary stats parts
update modules handling, add a function for finding the package path
the run module now handles input better
update options file behavior in edge cases
enforce guard clause indentation width
satisfy lint restrictions in cache helpers
silence ttl refresh replay noise
harden cache invalidation handling
drop stale memoise entries per invocation
use cli default handler fallback
remove non-ascii characters

Code Refactoring

unify significance marks option
fix the linear tests output
update the rest of the methods modules
finish the custom linear/nonlinear methods migration
fix lint issues and test suite issues for the custom methods
port legacy methods with regression tests
use cache runner for effect stats
update path handling
move module functions to a standalone folder
move the runtime method fetch function to a standalone modules folder
modularize options handling
add linters to tests, refactor fixtures
wrap all verbosity fetching into a single function
single out the function for printing options help text

Features

add non-linear publication bias diagnostics
allow configuring runtime method order
add linear regression diagnostics method
add reusable cache_cli runner wrappers
utilize the new caching system to cache data preprocessing
harden cache cli system
add a cli capture handler for print output
add new methods, fix caching, add optional cache disabling
add an explicit wrapper function for fetching cached artifacts
implement the custom caching functions and add test suites
add cache packages and test suites
add caching to the variable summary stats function
add a caching system
add an option to select how many decimal places to round the output to
control for verbosity in all of the existing cli logs
add the option to choose verbosity
implement the variable summary stats method

v0.2.1

2025-04-25

Bug Fixes

add missing explicit package imports back to the DESCRIPTION file
fix more data preprocessing import errors
fix an import error
add a couple of objects to the globals list

Code Refactoring

standardize the column names of required and expected variable names
rewrite data processing into a pipeline

Features

add valiadtion and value enforcement methods for data preprocessing
rename reg_df -> reg_dof
improve option parsing validation and type coercion
add a utility function for calculating study sizes
add a calc placeholder module and calculation for t-values
add methods for computing study/obs ID optional columns
in data processing, enforce the correct column types and names better
deprecate the column standardization
add further column validation to data preprocessing
add preprocessing methods for column name standardization
add the purrr package, add utility functions for parsing data
add a function for fetching definitions from option templates
add the ‘study’ column to the data recognized column names

v0.2.0

2025-04-24

Bug Fixes

fix the data config top level module import error
parsing option templates is now done through leaf paths
remove an unimported function
add a missing CONST import to the options utils module
add a missing argument to the main function
parsing the user options now properly displays help, modify the template to accommodate that
options now accept enum types

Code Refactoring

change the parsing of the options template to allow nested nodes
change the way the data config parsing function is structured
the ‘assert’ function now accepts a ‘msg’ argument

Features

partially enable data preprocessing, clean up cli messages
add functions for modifying, fixing, and fetching data configurations
add temporary options for better path awareness
implement a function to get a data config from the options file
prompts can now be defined through custom options
remove the unused data config methods - update through opt files now

BREAKING CHANGE

the new templates are now incompatible with old ones

v0.1.30

2025-04-16

Bug Fixes

add a missing import to the data config parsing function
add a missing export from the string module
rewrite the data config validation test to use the renamed function

Code Refactoring

rename the determine df type function (data_type -> df_type)
rename the data config file update function to better reflect its purpose
move the data config write function to a standalone module
decouple logic in data config writing/parsing
regex validation now happens against a more robust const object, add tests
rename the data config filename validation function
replace the internal ‘is_empty’ function with the in-built rlang function
remove the test constant and move the data into CONST.R

Features

add the body of data config parsing, add verbose string creation function
add various utility functions for parsing and writing the data config files
by default, the options copy now asks for overwrite permission in interactive mode
add functions for parsing data frames into data configs
add a function that validates an existence of a data config file
add a function for validating data config filenames
add a placeholder folder for data config files handling
remove the logger package, use ‘cli’ for all console logging

v0.1.29

2025-04-15

Bug Fixes

the cran submission workflow now parses the correct artifact name

v0.1.28

2025-04-15

v0.1.27

2025-04-15

v0.1.26

2025-04-15

v0.1.25

2025-04-15

Bug Fixes

add a missing newline to the release artifacts text file
the github workflow artifacts read now uses a more robust code
the build artifacts read workflow step now reads from a correct file
the gh release artifacts are now read from a single source of truth file
attempt to fix the build artifact upload path

v0.1.24

2025-04-14

Bug Fixes

the cran comments build workflow should now target the file correctly

v0.1.23

2025-04-14

Bug Fixes

change the release notes creation script to be more lenient
the update news action now updates with all the existing GH tags
replace the invalid slash in the git origin path in the news update workflow
remove the unused ‘replace’ argument in mock df creation
ArtmaBot now authenticates through the deploy key when updating the news file
the random vector generation now generates floats (was int) from a uniform distribution
use the official homebrew action to install homebrew
provide explicit pacakge method assignment when reading data
remove the homebrew install from the news update workflow
minor changes in the open PR script
comment out a PR open automation script irrelevant line
remove the mishandled ‘v’ tag
workflows to use a composite news creation action
more workflow issues
secret passing to the build workflow
the news update secrets passing
the cran submission workflow cli package call

Features

reading the data now validates the data frame as well

v0.1.22

2025-04-14

Code Refactoring

modify the build workflow to release from the CRAN version as opposed to the current one
the runtime module methods invocation function is now better formatted