Absolute risk from a sampled-cohort survival fit

Description

Estimates the cumulative incidence F_x(t) from a fitted survival design. Three engines are supported: the case-cohort pseudo-likelihood (“cch”, Borgan & Liestøl 1990) and the IPW nested case-control weighted Cox (“ipw_cox”, Samuelsen-weighted Breslow over the reused controls) both build F_x(t) = 1 − exp(−exp(β̂ᵀ x) Λ̂₀(t)) from an inverse-probability-weighted (IPW) Breslow cumulative baseline hazard Λ̂₀(t); the IPW nested case-control parametric AFT (“ipw_aft”, any of the Weibull / exponential / lognormal / loglogistic baselines) builds F_x(t) = G((log t − η̂)/σ̂) directly from the fitted survival curve, where η̂ is the AFT linear predictor, σ̂ the scale, and G the baseline error CDF. Pointwise confidence intervals use the delta method: for the Cox-type engines and the extreme-value AFT baselines this is the complementary log-log scale inverted to the risk scale; for the lognormal / loglogistic baselines it is the Wald interval on the standardised residual mapped through G.

Usage

absolute_risk(fit, ...)

## S3 method for class 'matchatr_fit'
absolute_risk(fit, newdata, times, conf_level = 0.95, ...)

Arguments

fit A matchatr_fit object returned by matcha(). The case-cohort (“cch”), IPW nested case-control weighted Cox (“ipw_cox”), and IPW nested case-control Weibull AFT (“ipw_aft”) engines are supported.
Unused; present for S3 consistency.
newdata A data frame of covariate values at which to evaluate F_x(t). Must contain columns matching the exposure and confounders used in fit. Each row is one covariate pattern; the result table carries a row column that indexes back to newdata.
times Non-empty numeric vector of evaluation times. Duplicates are dropped; times are sorted before evaluation. Times before the first event return F̂ = 0; times after the last event return the last Breslow value (step-function extrapolation).
conf_level Numeric in (0, 1). Confidence level for the pointwise intervals. Default 0.95.

Value

A matchatr_absolute_risk object. See absolute_risk.matchatr_fit for details on the return structure.

A matchatr_absolute_risk list with elements:

  • $estimates: a data.table with columns row (newdata row index), time, estimate (F̂_x(t)), ci_lower, ci_upper (delta-method CI on the probability scale).

  • $times: the sorted evaluation times.

  • $newdata: the supplied covariate patterns.

  • $conf_level, $ci_method, $engine, $estimator, $method.

References

Borgan O, Liestøl K (1990). A note on confidence intervals and bands for the survival function based on transformations. Scandinavian Journal of Statistics 17(1):35-41.

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

Kang S, Lu W, Liu M (2017). Efficient estimation for accelerated failure time model under case-cohort and nested case-control sampling. Biometrics 73(1):114-123.

See Also

matcha(), contrast(), case_cohort(), sample_ncc()

Other contrasts: contrast(), excess_risk(), print.matchatr_absolute_risk(), print.matchatr_excess_risk(), tidy.matchatr_absolute_risk(), tidy.matchatr_excess_risk()

Examples

library("matchatr")

fit <- matcha(cohort, outcome = "d", exposure = "x",
              design = case_cohort(subcohort = "sc", time = "t"),
              confounders = ~z, estimator = "cch")
absolute_risk(fit, newdata = data.frame(x = 1, z = 0), times = c(1, 2, 3))