BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Events - ECPv6.17.2//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:Events
X-ORIGINAL-URL:https://events.ucsc.edu
X-WR-CALDESC:Events for Events
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:20250309T100000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:20251102T090000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:20260308T100000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:20261101T090000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:20270314T100000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:20271107T090000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20260908T080000
DTEND;TZID=America/Los_Angeles:20260908T100000
DTSTAMP:20260902T163942Z
CREATED:20260902T163942Z
LAST-MODIFIED:20260902T163942Z
UID:10016350-1788854400-1788861600@events.ucsc.edu
SUMMARY:Zhang\, X. (STAT) - Predictive Generalized Variational Inference for Spatial Gaussian Process Model
DESCRIPTION:Spatially dependent data are common in environmental and climate studies\, where Gaussian process models are used for prediction and uncertainty quantification. Learning their covariance structure involves both statistical and computational challenges. Likelihood-based inference may be sensitive to covariance misspecification\, while repeated evaluation of prediction-oriented criteria can be computationally demanding. To study these issues\, we develop a predictive generalized variational inference (GVI) framework that learns a variational distribution by balancing proper-scoring-rule losses for conditional predictions with prior-based regularization. We first develop a marginal formulation for full Gaussian process models by analytically integrating out the regression coefficients and total residual variance. This formulation reduces the optimization dimension and focuses the variational optimization on the Matérn covariance parameters\, with Student-(t) conditional predictive distributions arising from the marginalization. We then develop a scalable formulation based on unordered nearest-neighbor conditionals. These conditionals define the predictive loss directly\, without treating their product as a joint likelihood. Because the same marginalization is not available in this formulation\, the regression coefficients and variance parameter are retained in a joint variational block. We evaluate logarithmic-score and continuous ranked probability score objectives through simulations and an analysis of December 2025 temperatures from 1\,218 U.S. Historical Climatology Network stations. The simulations compare different variational formulations\, scoring rules\, and covariance summaries using out-of-sample predictive evaluation. The real-data analysis compares likelihood-based and predictive GVI methods through cross-validation. We also examine the computational scaling of the nearest-neighbor formulation and consider extensions based on random local covariance matrices for heterogeneous spatial dependence\, as well as extensions to nonlinear regression and stochastic computer model emulation with quantitative and categorical inputs. Overall\, this work studies the specification and evaluation of predictive objectives\, uncertainty propagation\, and computational approximation within a unified variational framework. \nEvent Host: Xiao Zhang\, Ph.D. Student\, Statistical Science  \nAdvisor: Bruno Sansó \nZoom: https://ucsc.zoom.us/j/95501212725?pwd=WcB4SnkvhfoUFCm0Z4wUHLGB9omS05.1 \nPasscode: 697077
URL:https://events.ucsc.edu/event/zhang-x-stat-predictive-generalized-variational-inference-for-spatial-gaussian-process-model/
CATEGORIES:Ph.D. Presentations
ATTACH;FMTTYPE=image/jpeg:https://events.ucsc.edu/wp-content/uploads/2026/04/ph.d.-presentation-graphic-option-1.jpg
LOCATION:
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20260909T100000
DTEND;TZID=America/Los_Angeles:20260909T120000
DTSTAMP:20260819T162005Z
CREATED:20260819T162005Z
LAST-MODIFIED:20260819T162005Z
UID:10015347-1788948000-1788955200@events.ucsc.edu
SUMMARY:Chen\, Y. (STAT) - Flexible Bayesian Models for High-Dimensional and Longitudinal Discrete Data in Microbiome Studies
DESCRIPTION:Multivariate dependent discrete data routinely arise in microbiome studies. Analyzing these data presents interesting statistical challenges\, such as high dimensionality\, excess zeros\, large heterogeneity across samples\, and temporal dependence in longitudinal studies. Drawing inferences about objects of primary scientific interest—such as temporal trajectories of microbial abundance\, microbial interactions\, clusters of microbes similarly associated with environmental factors\, or networks of conditional dependencies—is more challenging due to these complexities\, requiring careful statistical modeling. Motivated by longitudinal microbiome experiments and large-scale fungal community surveys\, we propose flexible Bayesian models in which these objects are represented through low-dimensional or sparse structures\, regularized by global–local shrinkage priors\, and reported with uncertainty propagated through posterior inference. We first develop a Bayesian dynamic latent factor model for multivariate longitudinal count data. A rounded multivariate log-normal kernel links the observed counts to latent Gaussian variables\, and treatment-specific temporal trends of microbial abundance are modeled through Bayesian penalized B-splines. To better capture temporal changes in the mean\, dependence among microbial features is modeled through a low-dimensional factor structure with a Dirichlet–horseshoe+ shrinkage prior on the loadings. In addition\, continuous-time Ornstein–Uhlenbeck processes are used for the latent factors to account for temporal dependence within a subject. We next develop a sparse Bayesian probit regression model for high-dimensional presence–absence data to infer clusters of microbes whose presence has similar associations with environmental covariates. Due to the large number of microbes and excess zeros\, we first estimate a high-dimensional regression coefficient matrix using sparsity-inducing priors\, and then cluster the microbes based on the coefficient estimates. By using the posterior distribution of the coefficients\, we provide point estimates along with uncertainty quantification. Lastly\, we develop a Bayesian model that infers a time-varying precision matrix for longitudinal microbiome count data. While a covariance matrix describes marginal dependence\, a precision matrix characterizes conditional dependence among microbial features\, defining a microbial association network. By allowing this precision structure to evolve over time\, we obtain inferences about dynamic microbial networks. \nEvent Host: Yongqi Chen\, Ph.D. Student\, Statistical Science \nAdvisor: Juhee Lee \nZoom: https://ucsc.zoom.us/j/99307567941?pwd=NRnSLblnMKXEgqDRXPdBIFaRS8ctsq.1 \nPasscode: 301374
URL:https://events.ucsc.edu/event/chen-y-stat-flexible-bayesian-models-for-high-dimensional-and-longitudinal-discrete-data-in-microbiome-studies/
CATEGORIES:Ph.D. Presentations
ATTACH;FMTTYPE=image/jpeg:https://events.ucsc.edu/wp-content/uploads/2026/04/ph.d.-presentation-graphic-option2.jpg
LOCATION:
END:VEVENT
END:VCALENDAR