Statistics

Haruki Ebina discusses the marked power spectrum as a two-point approach to non-Gaussian density information, reducing the data-vector and covariance challenges of direct bispectrum analysis while retaining perturbative modelling. Recent work incorporates survey geometry, bringing applications to observations closer. Time permitting, he will also cover high-redshift DESI Run 2 cosmology and perturbative modelling of fast-radio-burst dispersion measures. Samuel Goldstein presents new constraints on fundamental physics from the cosmic microwave background and large-scale structure: axions using Atacama Cosmology Telescope observations, primordial non-Gaussianity, multi-field inflation using DESI, and component-separated temperature, polarization, thermal Sunyaev–Zeldovich and kinetic Sunyaev–Zeldovich maps from early Simons Observatory data. Speakers: Haruki Ebina, UC Berkeley, and Samuel Goldstein, Columbia University. Organized by Stanford KIPAC. Tuesday 29 September 2026, 10:40–11:30 America/Los_Angeles (PDT). Online via Zoom using Join the zoom meeting on the official event page. The organizer lists the intended audience as Faculty/Staff, Members and Students. Only remote attendance is verified for this listing.

Antonio Ocello tests the idea that heavy-tailed noise improves diffusion and flow-based generative models by matching heavy-tailed data and encouraging diverse samples. Replacing Gaussian noise also changes the estimation task. A combined theoretical and experimental study derives sampling-error bounds for representative heavy- and light-tailed diffusion models, finding that heavy-tailed noise makes statistical estimation harder and produces less favourable bounds. Experiments with synthetic and real data recover the predicted trade-off. The results question whether heavy-tailed initialization reliably improves exploration of rare regions. This is joint work with Hamza Cherkaoui and Hélène Halconruy. Online via Microsoft Teams. Friday 2 October 2026 at 15:30 CEST / 13:30 UTC / 14:30 BST (Europe/Paris). Follow the Microsoft Teams link beneath this occurrence on the seminar page; the public route offers browser or app access. Ocello is affiliated with ENSAE Paris. Organized by GAMEX — Generative AI Modeling for Extreme Events, with GLE²N and CIRCE support; the series welcomes scientific exchange beyond the network.

ISEC brings together ecologists, statisticians and quantitative researchers working on ecological inference and modelling. The main conference runs 10–15 January 2027 in Mérida, Mexico; separate pre-conference workshops run 8–9 January. Registration is open, with early rates through 30 September 2026 and reduced rates for eligible Global South participants.

AISTATS brings together machine learning, artificial intelligence and statistics, with an emphasis on methodological research. The 2027 conference is announced for 3–6 May in Montréal, Canada; the specific venue and registration details remain to be announced. The official schedule lists abstract submissions due 29 September 2026 and full papers due 6 October 2026, both at the end of the day Anywhere on Earth.

An undergraduate research fellowship examines air-temperature records across São Paulo state within a project on climate change and extreme-rainfall impacts. Tasks include gathering long meteorological time series, checking data quality, exploratory statistics, clustering, spatial interpolation and GIS mapping. The call is restricted to students already enrolled at FFLCH-USP who can devote at least 20 hours per week. The planned start is December 2026. Applications are accepted until 25 September 2026 using the email instructions on the investigator’s notice; applicants submit a Lattes CV, transcript and one-page statement of interest.

The Adolescent Accelerators Research Hub at UCT’s Centre for Social Science Research seeks a fellow to analyse quantitative evidence on adolescent health and interventions in resource-limited settings. Work includes analysis of longitudinal and cross-sectional data, methodological support, mentoring, data collection and publications, with collaborators at Oxford. The appointment runs from January 2027 to December 2028, with start-date flexibility. The fellowship is worth up to ZAR 350,000 annually, depending on experience and qualifications. Apply by 25 September 2026 with one PDF containing a suitability letter, CV and two academic referees, and a published quantitative writing sample. Follow the official call for submission details.

The Chair of Applied Macroeconomics and KOF Swiss Economic Institute invite applications for one or more doctoral positions using modern econometric methods and macroeconomic, firm-level, household, geospatial or international data. Research may examine globalisation and fiscal outcomes, multinational profit shifting and tax competition, international tax rules, political institutions and economic policy, or business cycles, forecasting, monetary policy, migration and corruption. Doctoral researchers will develop original papers, build and document datasets, present at seminars and conferences, join KOF research activities and complete the D-MTEC doctoral programme. The negotiable-start appointment is initially 18 months with possible extension through the normal doctorate; applications close 30 September 2026.

Develop and test methods for measuring public trust in policing, including approaches that reach communities underrepresented in existing research. The consortium will combine researchers, public partners and community expertise and evaluate practical interventions. The project must start by 1 February 2027 and last 36 months. UKRI covers 80% of project full economic costs up to £2 million. Applications close at 16:00 UK time on 27 October 2026.

Haruki Ebina discusses the marked power spectrum as a two-point approach to non-Gaussian density information, reducing the data-vector and covariance challenges of direct bispectrum analysis while retaining perturbative modelling. Recent work incorporates survey geometry, bringing applications to observations closer. Time permitting, he will also cover high-redshift DESI Run 2 cosmology and perturbative modelling of fast-radio-burst dispersion measures. Samuel Goldstein presents new constraints on fundamental physics from the cosmic microwave background and large-scale structure: axions using Atacama Cosmology Telescope observations, primordial non-Gaussianity, multi-field inflation using DESI, and component-separated temperature, polarization, thermal Sunyaev–Zeldovich and kinetic Sunyaev–Zeldovich maps from early Simons Observatory data. Speakers: Haruki Ebina, UC Berkeley, and Samuel Goldstein, Columbia University. Organized by Stanford KIPAC. Tuesday 29 September 2026, 10:40–11:30 America/Los_Angeles (PDT). Online via Zoom using Join the zoom meeting on the official event page. The organizer lists the intended audience as Faculty/Staff, Members and Students. Only remote attendance is verified for this listing.

Antonio Ocello tests the idea that heavy-tailed noise improves diffusion and flow-based generative models by matching heavy-tailed data and encouraging diverse samples. Replacing Gaussian noise also changes the estimation task. A combined theoretical and experimental study derives sampling-error bounds for representative heavy- and light-tailed diffusion models, finding that heavy-tailed noise makes statistical estimation harder and produces less favourable bounds. Experiments with synthetic and real data recover the predicted trade-off. The results question whether heavy-tailed initialization reliably improves exploration of rare regions. This is joint work with Hamza Cherkaoui and Hélène Halconruy. Online via Microsoft Teams. Friday 2 October 2026 at 15:30 CEST / 13:30 UTC / 14:30 BST (Europe/Paris). Follow the Microsoft Teams link beneath this occurrence on the seminar page; the public route offers browser or app access. Ocello is affiliated with ENSAE Paris. Organized by GAMEX — Generative AI Modeling for Extreme Events, with GLE²N and CIRCE support; the series welcomes scientific exchange beyond the network.

AHRC and DCMS invite teams to develop data-intensive methods for measuring participation at unticketed cultural, heritage and sporting settings. Proposals must demonstrate a usable measurement tool in at least one DCMS sector. The project must start on 1 May 2027 and finish by 31 March 2030. The full-economic-cost ceiling is £1.32 million; AHRC normally funds 80%. Applications opened on 22 September 2026 and close at 16:00 UK time on 1 December 2026.

AHRC and DCMS seek research evaluating the long-term consequences of major UK cultural and sporting events. The funded team will conduct a live experimental evaluation of a major event and improve methods for measuring its lasting effects. Projects must run from 1 May 2027 to no later than 31 March 2030. Full economic costs can reach £1.32 million, with AHRC normally paying 80%. Applications opened on 22 September 2026 and close at 16:00 UK time on 1 December 2026.