{"success":1,"msg":"","color":"rgb(28, 35, 49)","title":"A simple new approach to variable selection in regression, with application to genetic fine-mapping<\/b>","description":"webinar","title2":"","start":"2021-05-21 16:00","end":"2021-05-21 17:00","responsable":"Botond Szabo <\/i><\/a>","speaker":"Matthew Stephens (Chicago)","id":"42","type":"webinar","timezone":"Europe\/Amsterdam","activity":"Zoom meeting\r\n\r\nlink: https:\/\/vu-live.zoom.us\/j\/95319105586?pwd=MXlIWXIvNjcwQ0k1YzErMlkyTzNZQT09\r\nMeeting ID: 953 1910 5586\r\nPasscode: 645009\r\n\r\n","abstract":"We introduce a simple new approach to variable selection in linear\r\nregression, and to quantifying uncertainty in selected variables. The\r\napproach is based on a new model -- the\r\n``Sum of Single Effects'' (SuSiE) model -- which comes from writing\r\nthe sparse vector of regression coefficients as a sum of\r\n``single-effect'' vectors, each with one non-zero element. We also\r\nintroduce a corresponding new fitting procedure -- Iterative Bayesian\r\nStepwise Selection (IBSS) -- which is a Bayesian analogue of stepwise\r\nselection methods. IBSS shares the computational simplicity and speed\r\nof traditional stepwise methods, but instead of selecting a single\r\nvariable at each step, IBSS computes a {\\it distribution} on variables\r\nthat captures uncertainty in which variable to select.\r\nThe method leads to a convenient, novel, way to summarize uncertainty\r\nin variable selection, and provides a Credible Set for each selected\r\nvariable.\r\nOur methods are particularly well suited to settings where variables\r\nare highly correlated and true effects are sparse, both of which are\r\ncharacteristics of genetic fine-mapping applications.\r\nWe demonstrate through numerical experiments that our methods\r\noutperform existing methods for this task."}