BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Horizon Models With Dynamic Covariates: A Practitioner’s Perspec
 tive on Survival\, Reliability\, and the Limitations of Asymptotics
DTSTART;VALUE=DATE-TIME:20260325T150000Z
DTEND;VALUE=DATE-TIME:20260325T160000Z
DTSTAMP;VALUE=DATE-TIME:20260519T073800Z
UID:indico-event-2081@events.imath.kiev.ua
DESCRIPTION:Speakers: Igor  Fomenko (Keybank\, USA)\n\n\nThis talk does no
 t present new mathematical results. Instead\, it\nrevisits Oksana Chernova
 ’s work on Cox‑type models and baseline\nestimation—particularly the
  deep asymptotic theory and simulation\nevidence—from the standpoint of 
 a practitioner working across finance\nand biomarker research. Reflecting 
 on her results made me realize that\nit may be valuable to share my own un
 successful attempts to find\nrigorous mathematical support for the kinds o
 f Monte Carlo studies\nthat dominate both fields.\nAlthough the underlying
  questions appear similar\, the structural\ndifferences between domains ar
 e profound. In finance\, the number of\nloans is enormous\, observations a
 re nearly continuous\, and covariates\nevolve dynamically according to dif
 fusion‑type processes. In biomarker\nresearch\, sample sizes are small\,
  measurement times are sparse\, and\ncovariates are few and noisy. These d
 ifferences place the practitioner\nin a regime where classical asymptotic 
 arguments—so powerful in\nfixed‑design survival analysis—become surp
 risingly fragile.\nA central theme of the talk is why asymptotic foundatio
 ns remain\nelusive when covariates evolve dynamically over time\, despite 
 the\nenormous practical need for them. Once covariates follow stochastic\n
 processes\, the hazard becomes a random environment\, the cumulative\nhaza
 rd becomes a random functional\, and the likelihood no longer\nfactorizes 
 in a way that supports standard limit theory. The sample\nsize is “in th
 e wrong dimension”: finance has a large cross‑section but\nstrong temp
 oral dependence\, while biomarkers have weak dynamics but\ntiny cross‑se
 ction. As a result\, the model has no stable asymptotic\nobject to converg
 e to\, and the usual law‑of‑large‑numbers machinery\ncollapses. Yet 
 practitioners urgently need asymptotic anchors—what\nTukey called leadin
 g cases—to build toy models\, interpret simulations\,\nand create shared
  terminology.\n\nhttps://events.imath.kiev.ua/event/2081/
LOCATION:https://knu-ua.zoom.us/j/89643295643?pwd=eTBZZSt0d0thZzFyaUhDUFNG
 TVE3QT09 (ONLINE)
URL:https://events.imath.kiev.ua/event/2081/
END:VEVENT
END:VCALENDAR