Friday 2nd March 2012 – 12:45 to 13:45
Speaker: James Wolter (Yale)
I discuss estimation of hazard models where observations are dependent in three ways. First, covariates determining the hazard rate are allowed to be dependent. Covariates are either observation specific but allowed to be correlated or macro variables which all observations share. Second, all observations may be impacted by a global unobservable factor which varies with calendar time. This factor is referred to as frailty. Finally, the hazard rate of observations may be directly impacted by defaults of other related observations. This is given the generic name contagion.
Part of the OMI Seminar Series