MRTAnalysis 0.3.1
- Fixed a bug in
wcls() where input data with unordered
ID may cause dimension errors in matrix operations. Specifically, the
split() function was sorting IDs alphabetically while
cluster sizes and working covariance matrices used the order IDs
appeared in the data. Now uses factor() with explicit
levels to preserve ID ordering across all internal functions
(wcls_bread(), leverage(),
wcls_estfun(), wcls_meat(),
working.covariance()).
MRTAnalysis 0.3.0
- Added new functionality for mediated causal excursion effects in
MRTs:
- Added
mcee() function: streamlined workflow for
estimating natural direct excursion effect (NDEE) and natural indirect
excursion effect (NIEE) in micro-randomized trials (MRTs) with distal
outcomes.
- Added two advanced wrappers:
mcee_general(): flexible configuration of nuisance
models (p, q, eta, mu, nu) with support for multiple learners (glm, gam,
lm, rf, ranger, sl).
mcee_userfit_nuisance(): allows users to inject
externally fitted nuisance predictions.
- Included config helper functions (
mcee_config_glm(),
mcee_config_gam(), mcee_config_ranger(), etc.)
and mcee_config_maker() for building nuisance
specifications to pass into mcee_general().
- New dataset
data_time_varying_mediator_distal_outcome
included to illustrate usage.
- Added vignette “Time-Varying Causal Excursion Effect Mediation in
MRT: Continuous Distal Outcomes” with detailed examples and best
practices.
MRTAnalysis 0.2.0
- Added new functionality for distal outcomes in MRTs:
- Implemented
dcee() for estimating distal causal
excursion effects.
- Supports flexible nuisance regression learners (
lm,
gam, rf, ranger,
SuperLearner) with optional cross-fitting.
- Provides small-sample t inference via
summary.dcee_fit(), consistent with wcls() and
emee().
- New synthetic dataset
data_distal_continuous for
examples and testing.
- Added vignette: Exploratory Analysis for MRT: Distal Outcomes.
- Minor bug fixes and improvements to wcls() and emee()
documentation.
MRTAnalysis 0.1.2
- Fixed a bug in wcls when the randomization probability is
time-varying.
- Now all variable inputs need to be in quotation marks; for example,
from now on one should specify id = “userid” instead of id = userid.
This is to allow dynamically specified column names.
MRTAnalysis 0.1.1
- Updated vignette to improve clarify.
MRTAnalysis 0.1.0