CensoredDistributions.jl

Appearance

15

Release notes ​

CensoredDistributions.jl follows a hybrid changelog approach:

  • GitHub Releases: All releases (automatically generated by TagBot)

  • NEWS.md: Major releases requiring detailed context (shown below)

See GitHub Releases for minor/patch release information.

Ongoing releases are documented in GitHub releases,

v0.1.0 - Initial Release ​

CensoredDistributions.jl extends Distributions.jl to support primary event censoring and interval censoring, enabling modelling of scenarios where observation times are subject to various forms of censoring and truncation.

Core functionality ​

Distribution Constructors

  • primary_censored(dist, primary_event): Creates primary event censored distributions where an initiating event occurs within a time window, then experiences a delay

  • interval_censored(dist, interval): Creates interval censored distributions for continuous values observed only within discrete intervals (regular or arbitrary intervals)

  • double_interval_censored(dist; kwargs...): Combines primary event censoring, optional truncation, and optional secondary interval censoring in an appropriate mathematical sequence

Distribution Types

  • PrimaryCensored{D, P}: Wraps delay distribution D with primary event distribution P

  • IntervalCensored{D, T}: Wraps continuous distribution D with interval boundaries of type T

  • DoubleIntervalCensored: Composition of multiple censoring mechanisms

Extensible CDF Method

  • primarycensored_cdf(): User-extensible method for computing CDFs of primary censored distributions with analytical solutions for common distribution pairs

Utility Functions

  • weight(dist, weights): Creates weighted distributions for efficient likelihood computation

Distributions.jl interface ​

Partial implementation of Distributions.jl interface including:

  • pdf, logpdf: Probability density functions

  • cdf, logcdf: Cumulative distribution functions

  • quantile: Inverse CDF (where analytically tractable)

  • rand: Random number generation

  • minimum, maximum, insupport: Support queries

  • mean, var, std: Moments (where analytically tractable)

Migration from primarycensored R package ​

CensoredDistributions.jl provides a Julia-native implementation with enhanced functionality compared to the primarycensored R package:

Function Mapping

  • R's dprimarycensored() → Julia's pdf(primary_censored(...))

  • R's pprimarycensored() → Julia's cdf(primary_censored(...))

  • R's qprimarycensored() → Julia's quantile(primary_censored(...))

  • R's rprimarycensored() → Julia's rand(primary_censored(...))

Parameter Differences

  • Primary event window: Use primary_event = Uniform(0, pwindow) instead of R's pwindow parameter

  • Solver selection: Configure via solver keyword instead of R's integration method parameters

  • Truncation: Apply via truncated() from Distributions.jl or use double_interval_censored() convenience function

Contributors ​