NEWS

robust2sls 0.2.4 (2026-08-20)

This version only updates the unit tests and vignettes.

discovered slight changes in 14th or 15th decimals
discovered changes to some random number generations from generate_data(), which were traced back to changes in the environment and changes in the MASS pkg
restructured the tests to rely on fixtures
- created fixed datasets, stored in tests/testthat/testdata/
- replaced generate_data() with readRDS() where possible
- for mc_grid(), artifical data generation happens within the function body using generate_data(); to make it use fixed datasets, used the mockery pkg and macked the function to read in the stored datasets instead of drawing randim data
fixed GH issue #19, making parallel execution more robust
introduced some tolerance to numerical comparisons
- sup_test simulates pvalue, allowed for 0.015 tolerance diff (e.g. across OS)
- iis_init snapshots rounded to tolerance 12 digits, set to 0 if smaller than that

created two tags, which represented steps along the way of these improvements:
- v0.2.3-fixtures was developed on the old environment, changed tests to rely on the fixed datasets and ensured all tests still passed; then converted to new environment
- v0.2.3-ivreg-fixes updated code (tests and vignettes) to avoid new warnings about collinearity, which did not appear previously
- also updated the gets pkg to v0.40, no changes were required

move package doRNG from Imports to Suggests because the package has been orphaned as of 2022-12-19. Only affects the function mc_grid(), which is anyway for advanced users only

addition of Impulse Indicator Saturation (IIS) as initial estimator for outlier detection (from ivgets package)

code coverage has been extended to 100%
update formulae for FODR under normality (estimation not required)
all 2SLS estimation commands originally from AER package now use the ivreg package
fixed \eqn or \deqn or with empty second argument, as suggested by CRAN maintainers

Introduction of a suite of new functions implementing several tests for the presence of outliers in the sample.

A vignette is included that illustrates the usage of these testing functions.

change gauge_avar() to accept NULL as input for split argument
implement asymptotic covariance of the FODR for different cut-offs
utility function multi_cutoff() that allows to apply the same outlier detection algorithm with different cut-off / gamma values