BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//University of Iceland//AI Centre//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
UID:894d4d2d-8a73-4c82-bf8d-017f43999d8a@ai.hi.is
DTSTAMP:20261011T043452Z
DTSTART:20260127T090000Z
DTEND:20260127T100000Z
SUMMARY:Beyond Conditional Independence: Modeling Observation-Level Spatial
  Dependence in Hierarchical Models
DESCRIPTION:For normally distributed data\, modeling dependence is straight
 forward—we specify a covariance matrix. For non-Gaussian data\, hierarch
 ical models can introduce dependence on a latent scale through spatially c
 orrelated Gaussian fields\, but observations remain conditionally independ
 ent given the latent parameters. This ignores dependence in the data beyon
 d what the latent structure captures. In extreme rainfall modeling\, for i
 nstance\, nearby locations may produce observations with similar quantiles
  even after conditioning on their location\, scale and shape parameters—
 when rainfall at one site exceeds its median\, neighboring sites likely do
  too.\n \nI introduce a computationally tractable approximate Bayesian met
 hod that models this observation-level dependence using copulas. Simulatio
 ns show that ignoring such dependence biases parameter estimates\, while m
 odeling it improves point estimation. However\, an interesting trade-off e
 merges: the copula model can produce overly narrow parameter intervals whe
 n observation correlation is high\, even as observation-level intervals ma
 intain proper coverage\n\nhttps://ai.hi.is/is/events/colloquium-in-statist
 ics-and-ai-8/
LOCATION:Íslensk erfðagreining\, Tjarnarsalur
URL:https://ai.hi.is/is/events/colloquium-in-statistics-and-ai-8/
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR