One-Likelihood Conditional-Logistic Inference for KK Binary Designs

Estimates a treatment log-odds-ratio \(\beta_T\) for binary (incidence) outcomes collected under a KK matching-on-the-fly design (DesignSeqOneByOneKK14 or subclass) by maximizing one combined likelihood that couples a conditional-logistic (intercept-free, within-matched-pair) likelihood for matched subjects with an ordinary logistic likelihood for reservoir subjects, sharing a single treatment coefficient across both pieces. This is the "one-likelihood" counterpart to InferenceIncidKKCondLogitIVWC, which instead fits the matched and reservoir pieces separately and pools them by inverse-variance weighting; here the treatment coefficient is a single joint MLE, and likelihood_tier = "full" exposes likelihood-ratio, score, and gradient inference plus a parametric likelihood bootstrap in addition to Wald.

Details

Estimand. \(\beta_T\), the treatment coefficient of a logistic mean model \(\mathrm{logit}(P(Y=1 \mid w,x)) = \beta_0 + \beta_T w + x\beta\); \(\exp(\hat\beta_T)\) is the treatment-vs-control odds ratio.

Model. Matched pairs contribute McFadden-style conditional logistic likelihood terms that condition away the pair-specific nuisance intercept (see build_matching_combined_clogit_design_cpp/ collect_discordant_pairs_cpp); reservoir subjects contribute an ordinary logistic likelihood with one shared intercept. The combined negative log-likelihood is minimized jointly in \((\beta_0, \beta_T, \beta)\) via fast_logistic_regression_cpp/ fast_logistic_regression_with_var_cpp. When get_testing_type() != "wald", asymptotic CI/p-value calls are routed through InferenceAsympLik's generic score/likelihood-ratio/gradient dispatch instead of the design's own Wald machinery.

Assumptions. Independence across matched pairs and reservoir subjects given covariates; correct logistic mean specification; a KK matching-on-the-fly design supplying the matched/reservoir partition. No response censoring is supported (checked at construction via assertNoCensoring()).

References

Kapelner, A., and Krieger, A. M. (2014). "Matching on-the-fly: Sequential allocation with higher power and efficiency." Biometrics, 70(2), 378-388. doi:10.1111/biom.12148 . (KK14 in REFERENCES.md.)

See also

Analogous Python API for conditional logistic regression: statsmodels discrete models (ConditionalLogit). Logistic regression (orientation).

Super class

Inference -> InferenceIncidKKCondLogitOneLik

Methods

Public methods

• InferenceIncidKKCondLogitOneLik$compute_rand_two_sided_pval()

• InferenceIncidKKCondLogitOneLik$clone()

+ inherited public methods from Inference

• Inference$capabilities()
• Inference$compute_asymp_confidence_interval()
• Inference$compute_asymp_two_sided_pval()
• Inference$compute_estimate()
• Inference$compute_exact_confidence_interval()
• Inference$compute_exact_two_sided_pval_for_treatment_effect()
• Inference$duplicate()
• Inference$get_analysis_data()
• Inference$get_covariates()
• Inference$get_design_object()
• Inference$get_model_formula()
• Inference$get_nonestimable_reason()
• Inference$get_nonestimable_stage()
• Inference$get_optimization_alg()
• Inference$get_response()
• Inference$get_response_type()
• Inference$get_treatment()
• Inference$initialize()
• Inference$is_nonestimable()
• Inference$set_optimization_alg()
• Inference$set_seed()
• Inference$supports()

---

InferenceIncidKKCondLogitOneLik$compute_rand_two_sided_pval()

Usage

InferenceIncidKKCondLogitOneLik$compute_rand_two_sided_pval(
  r = 501,
  delta = 0,
  transform_responses = "none",
  na.rm = TRUE,
  show_progress = TRUE,
  permutations = NULL,
  type = NULL,
  args_for_type = NULL,
  zero_one_logit_clamp = .Machine$double.eps
)

---

InferenceIncidKKCondLogitOneLik$clone()

The objects of this class are cloneable with this method.

Usage

InferenceIncidKKCondLogitOneLik$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.