A Fixed Observational (Non-Randomized) Matched-Pair Design

An ObservationalDesign whose \(n\) subjects are organized into \(n/2\) matched pairs, each of size exactly 2. This is a convenience wrapper: it builds the canonical pair-membership vector \(m = (1, 1, 2, 2, \dots, n/2, n/2)\) internally – subject \(2k - 1\) and subject \(2k\) are pair \(k\) – and installs it via set_m(), so the caller need only supply subjects (and, later, responses) in that paired order; there is no separate m argument to get wrong.

Why not just ObservationalDesignBlocks with block size 2? It would look equivalent but silently behave differently: matching is a distinct capability (private$matching_capable, checked via is_matching_design()) from generic blocking, and several Inference components (jackknife, nonparametric bootstrap, Bayesian bootstrap, exchangeable-resampling-unit selection) branch on it to use pair-preserving resampling (MatchingStructure's draw_bootstrap_indices(), via draw_matching_bootstrap_sample_cpp()) instead of generic per-stratum resampling. ObservationalDesignBlocks overrides draw_bootstrap_indices() with the generic stratified version, so subclassing it here would advertise is_matching_design() == TRUE while still running the wrong bootstrap underneath. This class instead extends ObservationalDesign directly – the same relationship DesignFixedBinaryMatch has to DesignFixed – so it inherits MatchingStructure's matched-pair bootstrap machinery unmodified.

Super classes

Design -> DesignFixed -> ObservationalDesign -> ObservationalDesignMatching

Methods

Public methods

• ObservationalDesignMatching$new()

• ObservationalDesignMatching$clone()

+ inherited public methods from ObservationalDesign

• ObservationalDesign$assert_even_allocation()
• ObservationalDesign$supports_randomization_draw()
• ObservationalDesign$supports_resampling_replay()

+ inherited public methods from DesignFixed

• DesignFixed$add_all_subject_responses()
• DesignFixed$add_all_subjects_to_experiment()
• DesignFixed$assign_w_to_all_subjects()
• DesignFixed$overwrite_all_subject_assignments()

+ inherited public methods from Design

• Design$add_one_subject_response()
• Design$any_censoring()
• Design$applicable_inference_class_names()
• Design$assert_all_responses_recorded()
• Design$assert_all_subjects_arrived()
• Design$assert_fixed_sample()
• Design$capabilities()
• Design$check_experiment_completed()
• Design$draw_ws_according_to_design()
• Design$duplicate()
• Design$get_X()
• Design$get_X_imp()
• Design$get_X_raw()
• Design$get_design_formula()
• Design$get_edi_version_created()
• Design$get_effective_dead()
• Design$get_effective_time()
• Design$get_missingness_method()
• Design$get_n()
• Design$get_ordinal_levels()
• Design$get_original_ordinal_levels()
• Design$get_prob_T()
• Design$get_response_type()
• Design$get_response_type_original()
• Design$get_t()
• Design$get_w()
• Design$get_y()
• Design$get_y_L()
• Design$get_y_R()
• Design$get_y_original()
• Design$has_general_censoring()
• Design$incompatible_inference_classes_due_to_design_structure()
• Design$is_a_bernoulli_capable()
• Design$is_a_cluster_capable()
• Design$is_a_kk_matching_capable()
• Design$is_blocking_design()
• Design$is_fixed_sample_size()
• Design$is_matching_design()
• Design$prepare_for_resampling_replay()
• Design$randomization_family()
• Design$supports()
• Design$supports_resampling()
• Design$transform_y()
• Design$unavailable_inference_classes_due_to_missing_packages()
• Design$warm_all_subject_data_cache()

---

ObservationalDesignMatching$new()

Initialize a fixed observational (non-randomized) design whose subjects are organized into n / 2 matched pairs of size 2 (subjects 2k - 1 and 2k form pair k). Unlike ObservationalDesignBlocks there is no separate m argument — the pair structure is fixed by subject order and installed automatically via set_m(), and private$matching_capable is set so downstream Inference classes use pair-preserving (not generic stratified) resampling (see class documentation). As with every ObservationalDesign, \(w\) itself is supplied afterward via assign_w_to_all_subjects(w_precomputed = ...), in the same paired subject order.

Usage

ObservationalDesignMatching$new(
  response_type,
  n,
  include_is_missing_as_a_new_feature = TRUE,
  verbose = FALSE,
  missingness_method = "impute",
  design_formula = ~.,
  seed = NULL
)

Arguments

response_type

"continuous", "incidence", "proportion", "count", "survival", or "ordinal".

n

The sample size; must be even (subjects 2k - 1/2k form pair k, so an odd n would leave one subject unpaired).

include_is_missing_as_a_new_feature

Flag for missingness indicators.

verbose

A flag for verbosity.

missingness_method

How to handle missing values in covariates.

design_formula

A formula object.

seed

Integer seed for reproducibility.

Returns

A new `ObservationalDesignMatching` object

---

ObservationalDesignMatching$clone()

The objects of this class are cloneable with this method.

Usage

ObservationalDesignMatching$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

des = ObservationalDesignMatching$new(response_type = 'continuous', n = 6)
des$add_all_subjects_to_experiment(data.frame(x1 = rnorm(6)))
des$assign_w_to_all_subjects(w_precomputed = c(1, 0, 0, 1, 1, 0))