BEGIN:VCALENDAR PRODID:-//University of Illinois//Web Services Calendar//EN VERSION:2.0 CALSCALE:GREGORIAN BEGIN:VEVENT DTSTAMP:20120214T104046Z DTSTART;TZID=America/Chicago:20120216T160000 DTEND;TZID=America/Chicago:20120216T160000 SUMMARY:"Bayesian Clustering with Regression" by Peter Mueller\, Universi ty of Texas Austin CREATED:20110505T090000Z LAST-MODIFIED:20120127T100000Z LOCATION:165 Everitt Lab CATEGORIES:Seminar CONTACT:3-2167 ORGANIZER:office@illinois.edu URL:http://illinois.edu/calendar/detail/1439?key=20000101200001013824889 UID:3824889@illinois.edu END:VEVENT BEGIN:VEVENT DTSTAMP:20120214T104046Z DTSTART;TZID=America/Chicago:20120223T160000 DTEND;TZID=America/Chicago:20120223T160000 SUMMARY:UIUC-Purdue Joint Colloquium CREATED:20120117T100000Z DESCRIPTION:The study of prediction bias is important and the last five d ecades includes research studies that examined whether test scores diffe rentially predict academic or employment performance. Previous studies u sed ordinary least squares (OLS) to assess whether groups differ in inte rcepts and slopes. This study shows that OLS yields inaccurate inference s for prediction bias hypotheses. This paper builds upon the criterion-p redictor factor model by demonstrating the effect of selection\, measure ment error\, and measurement bias on prediction bias studies that use OL S. The range restricted\, criterion-predictor factor model is used to co mpute type I error and power rates associated with using regression to a ssess prediction bias hypotheses. In short\, OLS is not capable of testi ng hypotheses about group differences in latent intercepts and slopes. A dditionally\, a theorem is presented which shows that researchers should not employ hierarchical regression to assess intercept differences with selected samples. LAST-MODIFIED:20120119T100000Z LOCATION:Purdue University\, West Lafayette\, IN CATEGORIES:Seminar CONTACT:3-2167 ORGANIZER:office@illinois.edu URL:http://illinois.edu/calendar/detail/1439?key=200001012000010115066462 UID:15066462@illinois.edu END:VEVENT BEGIN:VEVENT DTSTAMP:20120214T104046Z DTSTART;TZID=America/Chicago:20120301T160000 DTEND;TZID=America/Chicago:20120301T160000 SUMMARY:David Ceperley\, Universitry of Illinois at Urbana-Champaign CREATED:20111109T100000Z LAST-MODIFIED:20120115T100000Z LOCATION:165 Everitt Lab CATEGORIES:Seminar CONTACT:3-2167 ORGANIZER:office@illinois.edu URL:http://illinois.edu/calendar/detail/1439?key=200001012000010112001231 UID:12001231@illinois.edu END:VEVENT BEGIN:VEVENT DTSTAMP:20120214T104046Z DTSTART;TZID=America/Chicago:20120308T160000 DTEND;TZID=America/Chicago:20120308T160000 SUMMARY:Han Xiao\, Rutgers CREATED:20120115T100000Z LAST-MODIFIED:20120115T100000Z LOCATION:165 Everitt Lab CATEGORIES:Seminar CONTACT:3-2167 ORGANIZER:office@illinois.edu URL:http://illinois.edu/calendar/detail/1439?key=200001012000010114958920 UID:14958920@illinois.edu END:VEVENT BEGIN:VEVENT DTSTAMP:20120214T104046Z DTSTART;TZID=America/Chicago:20120315T160000 DTEND;TZID=America/Chicago:20120315T160000 SUMMARY:Xiangrong Yin\, University of Georgia CREATED:20110817T090000Z LAST-MODIFIED:20120111T100000Z LOCATION:165 Everitt Lab CATEGORIES:Seminar CONTACT:3-2167 ORGANIZER:office@illinois.edu URL:http://illinois.edu/calendar/detail/1439?key=20000101200001018300163 UID:8300163@illinois.edu END:VEVENT BEGIN:VEVENT DTSTAMP:20120214T104046Z DTSTART;TZID=America/Chicago:20120405T160000 DTEND;TZID=America/Chicago:20120405T160000 SUMMARY:Hongyu Zhao\, Yale University CREATED:20110830T090000Z LAST-MODIFIED:20120111T100000Z LOCATION:165 Everitt Lab CATEGORIES:Seminar CONTACT:3-2167 ORGANIZER:office@illinois.edu URL:http://illinois.edu/calendar/detail/1439?key=20000101200001018899696 UID:8899696@illinois.edu END:VEVENT BEGIN:VEVENT DTSTAMP:20120214T104046Z DTSTART;TZID=America/Chicago:20120412T160000 DTEND;TZID=America/Chicago:20120412T160000 SUMMARY:Tucker McElroy\, U.S. Census Bureau CREATED:20111231T100000Z LAST-MODIFIED:20120111T100000Z LOCATION:165 Everitt Lab CATEGORIES:Seminar CONTACT:3-2167 ORGANIZER:office@illinois.edu URL:http://illinois.edu/calendar/detail/1439?key=200001012000010114275620 UID:14275620@illinois.edu END:VEVENT BEGIN:VEVENT DTSTAMP:20120214T104046Z DTSTART;TZID=America/Chicago:20120419T160000 DTEND;TZID=America/Chicago:20120419T160000 SUMMARY:Tong Zhang\, Rutgers CREATED:20111201T100000Z LAST-MODIFIED:20120116T100000Z LOCATION:165 Everitt Lab CATEGORIES:Seminar CONTACT:3-2167 ORGANIZER:office@illinois.edu URL:http://illinois.edu/calendar/detail/1439?key=200001012000010112977173 UID:12977173@illinois.edu END:VEVENT BEGIN:VEVENT DTSTAMP:20120214T104046Z DTSTART;TZID=America/Chicago:20120426T160000 DTEND;TZID=America/Chicago:20120426T160000 SUMMARY:"High Dimensional Statistical Learning with Bayesian Variable Sel ection" by Wenxin Jiang\, Northwestern University CREATED:20120116T100000Z DESCRIPTION:This is a theoretical study on the frequentists' convergence properties of Bayesian inference using binary logistic or probit regress ion\, when the number of explanatory variables `p' is possibly much larg er than the number of study units `n'.In a popular approach of `Bayesian Variable Selection'\, one uses a prior to select a limited number of ca ndidate variables to enter the model. We show that this approach can ind uce posterior estimates of the regression functions that are consistentl y estimating the truth\, if the true regression model satisfies some `sp arseness condition'\, which indicates that most of the candidate variabl es have very small effects in regression. The estimated regression funct ions therefore can also produce `consistent' classifiers that are asympt otically optimal for predicting future binary responses. Furthermore\, w e show in some sparse situations that the corresponding rate of converge nce resembles the convergence rate in a lowdimensional setup (p << n)\, even if the actual set up is high dimensional with p >> n. Therefore\, i t is possible to use Bayesian variable selection to reduceoverfitting ca used by the curse of dimensionality. LAST-MODIFIED:20120116T100000Z LOCATION:165 Everitt Lab CATEGORIES:Seminar CONTACT:3-2167 ORGANIZER:office@illinois.edu URL:http://illinois.edu/calendar/detail/1439?key=200001012000010115006785 UID:15006785@illinois.edu END:VEVENT END:VCALENDAR