Every job template the package ships, rendered once on the same synthetic study. Each row says what the template is for and gives the add_job() call that adds it to your study; copy it, change subject and type to name your job, and run it from inside the study. Open a template's name to see the report it produces.
Naming a job: subject and type
add_job() needs three things. The first argument picks the template. subject and type are names you choose: there is no list to pick from, only the rule that they use letters, digits and underscores.
They name the job file, <subject>-<type>-<template>.qmd, and the folder its results are saved in.
Jobs in one analysis must share them. hm finds the hz.rds that hz saved only when both carry the same subject and type, so a hazard analysis of death might use subject = "death", type = "hz" for ac, hz, hm and hp.
The calls below carry each template's own example values, which this gallery's study also used. Change them to name your analysis. The add_job() help page has the details.
The study
800 simulated operations, 1990 to 2024, followed to 31 December 2025. No real patient, study or identifier appears anywhere.
Built from the descriptives demo cohort, plus the outcomes each family needs: a two-phase death outcome (early and constant hazard, concordance 0.69), ventilation, ICU hours, stroke, regurgitation grade, discharge, and three treatment variables, plus a separate synthetic echocardiography series for the repeated-measures job.
Each job was scaffolded with add_job(), its EDIT: points filled, and rendered with render_job(final = TRUE).
Where the gallery departs from the template
Deaths on the day of operation are recorded at 0.00025 years in the dataset itself, since the hazard likelihood is undefined at zero (#175).
The hazard chain analyses complete cases on creatinine, a WHERE on ac and hz that hm, hp and hs-setup inherit from hz.rds (#174).
Each bootstrap bag is made by the report's own shipped runner, run small: bh 8 replicates in four chunks with its early shape held fixed, the others 100, so frequencies are rough.
lm-binary trains on operations before 2015 and lm-checkpred validates on 2015 onward, by each job's WHERE (#180).
hs-concordance's two group models hold the early-phase nu fixed at the whole-cohort estimate and screen three candidates; free, the transcatheter group's early phase collapses onto its four same-day deaths.