[Forum SIS] Prof. Sunil J. Rao - Seminari su Precision Medicine e Shrinkage estimation

[Claudio Conversano]

Si comunica che il prof. Sunil J. Rao (Professor and Director, Division of Biostatistics Miller School of Medicine University of Miami, USA) 
https://umiamihealth.org/sylvester-comprehensive-cancer-center/research/faculty/j-sunil-rao-phd
 terrà I seguenti seminari presso l'università di Cagliari:

1) Some Statistical Adventures in Precision Medicine
Lunedì 21 ottobre, ore 17.00, Aula 15, Blocco I (ex Asse E), Cittadella Universitaria (Monserrato)

2) Shrinkage Estimation for Linear Models - Something Old Becomes Something New
Venerdì 25 ottobre, ore 11.00, Aula Informatica, Edificio Baffi, Dipartimento di Scienze Economiche e Aziendali, Università di Cagliari, Vial Frà Ignazio

Si riportano di seguito gli abstract dei due seminari e un breve profilo del relatore.

Tutti gli interessati sono invitati a partecipare.

Per informazioni:
Claudio Conversano
Dipartimento di Scienze Economiche e Aziendali
Viale Frà Ignazio 17, 09123 Cagliari
conversa a unica.it
070-6753337

1) Some Statistical Adventures in Precision Medicine

Abstract: Precision medicine has the potential to revolutionize the way patients are treated for complex diseases like cancer with the promise of improved responses to treatments, coupled with reduced unwanted side effects.  
In fact, precision medicine is widely advertised as a treatment option strategy at many medical centers in the US today.  However, there are some very interesting and open statistical issues that the precision medicine paradigm reveals and in fact are necessary to solve in order to make the paradigm operational. In a sense, because the field is so nascent, it’s a bit like being on an adventure looking for a statistical path to safety.   
In this talk, I will describe three vignettes where we have developed novel statistical methods to some of these open problems. The vignettes will cover pharmacogenomic experiments in cancer, validation of findings from a high throughput precision medicine experiment, and the problem of accurate prediction of outcomes.

2) Shrinkage Estimation for Linear Models - Something Old Becomes Something New

Abstract: Shrinkage estimation for linear models has been around for many years but experienced a significant resurgence with the dawn of the high dimensional data (e.g. genomic data).  As a result, there may be no other area in statistics with as much attention in recent literature than shrinkage estimation.    I will go back in time and talk about an original motivation for shrinkage - ridge regression to address collinearity in the predictors.  I will then describe how ridge regression can be extended to allow a wider class of shrinkage estimators which includes the highly popular lasso and elastic net methods.  The lasso was developed by Rob Tibshirani while I was his student many years ago.  Finally, I will talk about some of my own research in this area called spike and slab regression which is a particular implementation of generalized ridge regression.  Throughout, I will demonstrate the usefulness of these methods on real data analyses and talk a little about accompanying R software.

Breve profilo biografico del relatore
Dr. J. Sunil Rao is the Interim Chair of the Department of Public Health Sciences (2016+) and the Director of the Division of Biostatistics at the University of Miami Miller School of Medicine (2010+).  [[During his tenure as Chair, the Department’s research portfolio has grown from $10 USD million per year to over $20 USD million per year]].  Prior to this, he was Director of the Division of Biostatistics at Case Western Reserve University in Cleveland, Ohio (1998-2010).  He received his Ph.D. under the supervision of Dr. Robert Tibshirani at the University of Toronto (1994) focusing on bootstrap model selection techniques.
His areas of methodological research include Bayesian model selection using spike and slab priors, mixed model selection, mixed model prediction, subgroup identification methods, small area estimation and pharmacogenomic modeling.  Specifically recognized advances include spike and slab regression, the observed best predictor, fence methods for mixed model selection, the E-MS algorithm, classified mixed model prediction and local sparse bump hunting.  His primary areas of application are currently in cancer genomics, drug repositioning using electronic health record data and the analysis of health disparity data.  
Dr. Rao is a Fellow of the American Statistical Association and Elected Member of the International Statistical Institute.