BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//juliacon-2026//talk//ZUMSFD
BEGIN:VTIMEZONE
TZID:Europe/Berlin
BEGIN:DAYLIGHT
DTSTART:20250814T000000
TZNAME:CEST
TZOFFSETFROM:+0200
TZOFFSETTO:+0200
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20251026T030000
RDATE:20261025T030000
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20260329T030000
RDATE:20270328T030000
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:Estimating epidemiological delay distributions: from R/Stan to Jul
 ia - Sam Abbott
DTSTART;TZID=Europe/Berlin:20260814T144500
DTEND;TZID=Europe/Berlin:20260814T150000
DTSTAMP:20260804T031039Z
UID:pretalx-juliacon-2026-ZUMSFD@pretalx.com
DESCRIPTION:Delay distributions describe the time between epidemiological 
 events\, such as infection to symptom onset or symptom onset to hospitalis
 ation.\nEstimating these distributions from outbreak data is difficult bec
 ause both the primary event (e.g. infection) and the secondary event (e.g.
  symptom onset) are usually only known to have occurred within a time wind
 ow\, such as a day.\nReal-time outbreak data is also often right-truncated
  as longer delays have not yet been observed.\nIgnoring double interval ce
 nsoring and truncation biases parameter estimates which are then used for 
 forecasting and transmission modelling.\n\nAdjusting distributions for pri
 mary event censoring addresses this by integrating the delay CDF over the 
 primary event window\, weighted by the density of when\, within the window
 \, the event occurred.\nThis can then be combined with truncation and seco
 ndary interval-censoring adjustments to produce a double-interval-censored
  and right-truncation-adjusted distribution.\n\nIn this talk\, we present 
 [CensoredDistributions.jl](https://censoreddistributions.epiaware.org)\, w
 hich implements these adjustments as `primary_censored`\, `interval_censor
 ed`\, and `double_interval_censored`\, composable [Distributions.jl](https
 ://github.com/JuliaStats/Distributions.jl) wrappers.\nMultiple dispatch se
 lects closed-form CDFs for delay and primary event distribution pairs wher
 e these are available\, and falls back to numerical integration otherwise.
 \nWe demo the package standalone and with [Turing.jl](https://turinglang.o
 rg/) for parameter estimation.\n\nWe then compare to [primarycensored](htt
 ps://primarycensored.epinowcast.org)\, our equivalent R package\, which al
 so ships a duplicate set of [Stan](https://mc-stan.org/) functions so user
 s can fit models in either language.\nMaintaining two parallel implementat
 ions required reimplementing distribution functions in Stan\, building too
 ling to vendor Stan code into downstream projects\, and replacing types wi
 th integer distribution identifiers.\nStan's integral solver was also unst
 able for this problem\, so we had to recast it as an ODE.\nJulia's multipl
 e dispatch and ecosystem composability eliminates all of this.\n\nWe then 
 summarise our plans to build a composed Julia version of our [epidist](htt
 ps://epidist.epinowcast.org) R package\, using CensoredDistributions.jl as
  a foundation with Turing.jl submodels for partially pooled and flexible d
 elay estimation.
LOCATION:Muschel — N2
URL:https://pretalx.com/juliacon-2026/talk/ZUMSFD/
END:VEVENT
END:VCALENDAR