KK GLMM Inference for Proportion Responses

Fits a combined conditional-logit-plus-random-intercept-GLMM likelihood for proportion responses under a KK matching-on-the-fly design. Matched pairs with a discordant pair-difference are handled by a conditional (fixed pair-effect) logistic term, while concordant/reservoir subjects are handled by a random-intercept logistic mixed model with intercept \(b_g \sim N(0, \sigma_b^2)\) per matched-set/reservoir group \(g\); both terms share the same treatment coefficient \(\beta_T\), jointly maximized by the internal fast_clogit_plus_glmm_cpp routine. This combines the design-exact conditional-logit treatment of matched pairs (no nuisance pair-intercept to estimate) with a GLMM's ability to still contribute information from concordant pairs and reservoir subjects, which a pure conditional-logit-on-discordant-pairs-only approach would discard. \(\exp(\hat\beta_T)\) is the common treatment odds ratio. likelihood_tier = "full": likelihood-ratio, score, and Wald tests are all available when the model converges. See InferenceAbstractKKCondLogitGLMM for the shared model-fitting and caching contract used by this class's incidence-response siblings (InferenceIncidKKCondLogitGLMMIVWC, InferenceIncidKKCondLogitGLMMOneLik).

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 , for the KK matching-on-the-fly design this class is built for; Breslow, N. E., and Clayton, D. G. (1993). "Approximate Inference in Generalized Linear Mixed Models." Journal of the American Statistical Association, 88(421), 9-25, doi:10.2307/2290687 , for the GLMM likelihood framework combined with the conditional-logit term here.

Super classes

Inference -> InferenceAbstractKKCondLogitGLMM -> InferencePropKKGLMM

Methods

Public methods

• InferencePropKKGLMM$new()

• InferencePropKKGLMM$clone()

+ inherited public methods from InferenceAbstractKKCondLogitGLMM

• InferenceAbstractKKCondLogitGLMM$approximate_bayesian_bootstrap_distribution_beta_hat_T()
• InferenceAbstractKKCondLogitGLMM$approximate_bootstrap_distribution_beta_hat_T()
• InferenceAbstractKKCondLogitGLMM$approximate_jackknife_distribution_beta_hat_T()
• InferenceAbstractKKCondLogitGLMM$approximate_m_out_of_n_bootstrap_distribution_beta_hat_T()
• InferenceAbstractKKCondLogitGLMM$approximate_rand_bootstrap_distribution_beta_hat_T()
• InferenceAbstractKKCondLogitGLMM$approximate_randomization_distribution_beta_hat_T()
• InferenceAbstractKKCondLogitGLMM$approximate_subsampling_distribution_beta_hat_T()
• InferenceAbstractKKCondLogitGLMM$compute_asymp_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_asymp_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_bayesian_bootstrap_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_bayesian_bootstrap_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_bootstrap_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_bootstrap_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_estimate()
• InferenceAbstractKKCondLogitGLMM$compute_estimate_with_bootstrap_weights()
• InferenceAbstractKKCondLogitGLMM$compute_gradient_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_gradient_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_jackknife_bias_estimate()
• InferenceAbstractKKCondLogitGLMM$compute_jackknife_estimate()
• InferenceAbstractKKCondLogitGLMM$compute_jackknife_std_error()
• InferenceAbstractKKCondLogitGLMM$compute_jackknife_wald_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_jackknife_wald_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_lik_ratio_bartlett_approx_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_lik_ratio_bartlett_approx_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_lik_ratio_bartlett_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_lik_ratio_bartlett_exact_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_lik_ratio_bartlett_exact_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_lik_ratio_bartlett_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_lik_ratio_bootstrap_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_lik_ratio_bootstrap_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_lik_ratio_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_lik_ratio_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_m_out_of_n_bootstrap_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_m_out_of_n_bootstrap_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_param_bootstrap_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_param_bootstrap_estimate()
• InferenceAbstractKKCondLogitGLMM$compute_param_bootstrap_pval()
• InferenceAbstractKKCondLogitGLMM$compute_rand_bootstrap_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_rand_bootstrap_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_rand_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_rand_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_score_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_score_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_subsampling_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_subsampling_sensitivity()
• InferenceAbstractKKCondLogitGLMM$compute_subsampling_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$compute_wald_confidence_interval()
• InferenceAbstractKKCondLogitGLMM$compute_wald_two_sided_pval()
• InferenceAbstractKKCondLogitGLMM$get_information_preference()
• InferenceAbstractKKCondLogitGLMM$get_information_source_used()
• InferenceAbstractKKCondLogitGLMM$get_last_param_bootstrap_diagnostics()
• InferenceAbstractKKCondLogitGLMM$get_last_param_bootstrap_estimate_diagnostics()
• InferenceAbstractKKCondLogitGLMM$get_mod()
• InferenceAbstractKKCondLogitGLMM$get_summary()
• InferenceAbstractKKCondLogitGLMM$get_supported_bayesian_bootstrap_ci_types()
• InferenceAbstractKKCondLogitGLMM$get_supported_bayesian_bootstrap_pval_types()
• InferenceAbstractKKCondLogitGLMM$get_supported_bootstrap_ci_types()
• InferenceAbstractKKCondLogitGLMM$get_supported_bootstrap_pval_types()
• InferenceAbstractKKCondLogitGLMM$get_supported_information_preferences()
• InferenceAbstractKKCondLogitGLMM$get_supported_rand_bootstrap_ci_types()
• InferenceAbstractKKCondLogitGLMM$get_supported_rand_bootstrap_pval_types()
• InferenceAbstractKKCondLogitGLMM$get_supported_testing_types()
• InferenceAbstractKKCondLogitGLMM$get_testing_type()
• InferenceAbstractKKCondLogitGLMM$select_optimal_b_subsampling()
• InferenceAbstractKKCondLogitGLMM$select_optimal_m_out_of_n_bootstrap()
• InferenceAbstractKKCondLogitGLMM$set_custom_randomization_statistic_cpp()
• InferenceAbstractKKCondLogitGLMM$set_custom_randomization_statistic_function()
• InferenceAbstractKKCondLogitGLMM$set_information_preference()
• InferenceAbstractKKCondLogitGLMM$set_testing_type()
• InferenceAbstractKKCondLogitGLMM$supports_rand_pval_for_incidence()

+ inherited public methods from Inference

• Inference$capabilities()
• Inference$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$is_nonestimable()
• Inference$set_optimization_alg()
• Inference$set_seed()
• Inference$supports()

---

InferencePropKKGLMM$new()

Initialize inference for the combined conditional-logit (discordant matched pairs) plus random-intercept-GLMM (concordant pairs/reservoir) proportion model; see InferencePropKKGLMM for the model form. Does not fit the model; the fit is deferred to the first call to compute_estimate() or a method that requires it.

Usage

InferencePropKKGLMM$new(
  des_obj,
  model_formula = NULL,
  max_abs_reasonable_coef = 10000,
  max_abs_log_sigma = 8,
  verbose = FALSE,
  smart_cold_start_default = NULL,
  optimization_alg = NULL
)

Arguments

des_obj

A completed Design object with a proportion response.

model_formula

Optional formula for covariate adjustment.

max_abs_reasonable_coef

Cap for reasonable coefficient estimates.

max_abs_log_sigma

Cap for reasonable log random effect variance.

verbose

Whether to print progress messages.

smart_cold_start_default

Whether to use smart cold start values.

optimization_alg

Character. Optimization algorithm (default "lbfgs").

---

InferencePropKKGLMM$clone()

The objects of this class are cloneable with this method.

Usage

InferencePropKKGLMM$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.