NEWS.md
In NonlinearDiD: Staggered Difference-in-Differences with Nonlinear Outcomes

NonlinearDiD 0.1.0

nonlinear_attgt(): Core estimator for group-time ATT(g,t) under logit, probit, Poisson, negative binomial, and linear outcome models with staggered treatment adoption.
nonlinear_aggte(): Aggregation into event-study (dynamic), group-level, calendar-time, and overall ATT estimates.
nonlinear_pretest(): Pre-treatment parallel trends test with joint chi-squared test and HonestDiD-style sensitivity analysis.
binary_did_logit() / binary_did_probit(): Simple 2×2 DiD with binary outcomes on the log-odds / probit scale, with APE reporting.
binary_did_dr(): Doubly-robust binary DiD combining logit outcome regression with inverse probability weighting.
count_did_poisson(): Poisson QMLE DiD for count outcomes following Wooldridge (2023), reporting rate ratios.
odds_ratio_did(): Odds-ratio DiD estimator (scale-free, symmetric).
nonlinear_bounds(): Nonparametric Manski bounds and PT-restricted bounds for binary outcomes.
sim_binary_panel() / sim_count_panel(): Data-generating processes for simulation studies with staggered treatment and heterogeneous effects.
S3 methods: print(), summary(), plot() for all main object classes.
Rcpp-accelerated bootstrap weight generation and DR score computation.

This is version 0.1.0 — an initial implementation of a methodology that is actively being developed in the econometrics literature. The core identification arguments follow Roth & Sant'Anna (2023) and Wooldridge (2023). Standard errors are based on influence function / sandwich estimators.

Known limitations: - Simultaneous confidence bands use a conservative normal approximation; exact bands require the multiplier bootstrap (boot = TRUE). - Negative binomial staggered DiD uses approximation; full MLE version is planned for v0.2.0.