BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Drupal iCal API//EN X-WR-CALNAME:Events items teaser X-WR-TIMEZONE:America/Toronto BEGIN:VTIMEZONE TZID:America/Toronto X-LIC-LOCATION:America/Toronto BEGIN:DAYLIGHT TZNAME:EDT TZOFFSETFROM:-0500 TZOFFSETTO:-0400 DTSTART:20200308T070000 END:DAYLIGHT BEGIN:STANDARD TZNAME:EST TZOFFSETFROM:-0400 TZOFFSETTO:-0500 DTSTART:20191103T060000 END:STANDARD END:VTIMEZONE BEGIN:VEVENT UID:682a46be448e5 DTSTART;TZID=America/Toronto:20200821T103000 SEQUENCE:0 TRANSP:TRANSPARENT DTEND;TZID=America/Toronto:20200821T113000 URL:/statistics-and-actuarial-science/events/department -seminar-martin-larsson-carnegie-mellon-university SUMMARY:Department Seminar by Martin Larsson\, Carnegie Mellon University CLASS:PUBLIC DESCRIPTION:Summary \n\nOPTIMAL SUPERMARTINGALES FOR ANYTIME-VALID SEQUENTI AL TESTING\n\n-------------------------\n\nStatistical testing is`anytime- valid’ if the decision to stop or\ncontinue an experiment can depend on anything that has been observed\nso far\, without compromising statistical error guarantees. For\ninstance\, suppose that a promising but inconclusi ve study receives\nfunding to gather additional data. Then standard p-valu e analysis is\ninvalidated\, but anytime-valid testing is not. A recent ap proach to\nanytime-valid testing views a test statistic as a bet against t he null\nhypothesis. These bets are constrained to be supermartingales - h ence\nunprofitable - under the null\, but designed to be profitable under the\nrelevant alternative hypotheses. This perspective opens the door to\n using tools from financial mathematics. In this talk I will explain\nhow n otions such as supermartingale measures\, fork-convexity\, the\noptional d ecomposition theorem\, and universal portfolios can be used\nto design opt imal supermartingales for anytime-valid sequential\ntesting. (This talk is based on ongoing work with Aaditya Ramdas (CMU)\nand Johannes Ruf (LSE).)  \n\nPLEASE NOTE: This talk will be given through Webex. To join\, pleas e\nclick here: Department seminar by Martin Larsson\n[https://uwaterloo.w ebex.com/uwaterloo/onstage/g.php?MTID=e9e0f4bdb17f4e6e07b8c3abb701fd4e4]\n DTSTAMP:20250518T204446Z END:VEVENT END:VCALENDAR