Case-cohort design

Description

Declares a case-cohort sample: a random subcohort drawn from the full cohort at baseline, augmented by every case occurring in the cohort (including cases outside the subcohort). The design-faithful analyses are the Prentice, Self-Prentice, and Borgan pseudo-likelihood Cox estimators (survival::cch), whose controls are reused across failure times so the variance comes from a robust / asymptotic correction, not the information matrix.

Usage

case_cohort(subcohort, time, method = "Prentice", id = NULL, stratum = NULL)

Arguments

subcohort A single character string naming the 0/1 (or logical) column flagging subcohort membership.
time A single character string naming the event/follow-up time column.
method Character scalar naming the pseudo-likelihood method passed to survival::cch(). One of “Prentice” (the default), “SelfPrentice”, “LinYing”, “I.Borgan”, or “II.Borgan”. The first three are for simple (unstratified) subcohorts; the Borgan variants require stratified subcohort sampling and need stratum to be supplied. “Prentice” and “SelfPrentice” share the same asymptotic variance; “LinYing” uses an independent variance estimator; “I.Borgan” and “II.Borgan” are IPW estimators for stratified subcohorts with plug-in asymptotic variance.
id NULL or a single character string naming the subject-identifier column. When NULL (the default) the original row positions in the data are used as synthetic IDs. Supply an ID column when subjects can appear both as subcohort members and as cases (the common case-cohort situation): survival::cch() uses the ID to correctly pair each subject’s two appearances.
stratum NULL or a non-empty character vector naming the column(s) defining the subcohort sampling strata. Required when method is “I.Borgan” or “II.Borgan”: both IPW estimators weight each subject by the inverse of its stratum-specific subcohort sampling fraction, so the stratum boundaries must be known. When method is “Prentice”, “SelfPrentice”, or “LinYing” the stratum argument is ignored (those estimators assume a simple random subcohort).

Value

A matchatr_design object of type “case_cohort” carrying the subcohort, time, method, stratum, and (optionally) id columns, with a weight_spec flagged for inclusion-probability weighting.

References

Prentice RL (1986). A case-cohort design for epidemiologic cohort studies and disease prevention trials. Biometrika 73(1):1-11.

Self SG, Prentice RL (1988). Asymptotic distribution theory and efficiency results for case-cohort studies. The Annals of Statistics 16(1):64-81.

Lin DY, Ying Z (1993). Cox regression with incomplete covariate measurements. Journal of the American Statistical Association 88(424):1341-1349.

Borgan O, Langholz B, Samuelsen SO, Goldstein L, Pogoda J (2000). Exposure stratified case-cohort designs. Lifetime Data Analysis 6(1):39-58.

See Also

nested_cc(), matcha()

Other design constructors: counter_matched(), matched_cc(), nested_cc(), two_phase(), unmatched_cc()

Examples

library("matchatr")

case_cohort(subcohort = "in_subcohort", time = "t")
<matchatr_design>
 Type:       Case-cohort
 Time:       t
 Subcohort:  in_subcohort
 Method:     Prentice
 Weights:    inclusion-probability
case_cohort(subcohort = "in_subcohort", time = "t", method = "LinYing")
<matchatr_design>
 Type:       Case-cohort
 Time:       t
 Subcohort:  in_subcohort
 Method:     LinYing
 Weights:    inclusion-probability
case_cohort(subcohort = "in_subcohort", time = "t", id = "subject_id")
<matchatr_design>
 Type:       Case-cohort
 Time:       t
 Subcohort:  in_subcohort
 Method:     Prentice
 Weights:    inclusion-probability
case_cohort(subcohort = "in_subcohort", time = "t",
            method = "I.Borgan", stratum = "region")
<matchatr_design>
 Type:       Case-cohort
 Time:       t
 Subcohort:  in_subcohort
 Method:     I.Borgan
 Stratum:    region
 Weights:    inclusion-probability