Joint Model For Complex Clinical Data

Complex clinical data such as AIDS clinical data usually contains correlated longitudinal data(Longitudinal Data)and survival data(time-to-Event Data),respectively,the longitudinal and survival of the data modeling method because of the neglect of this nature of the association often lead to large Tolerance(Guo 2004).A large number of literatures have shown that the joint modeling of complex clinical data such as AIDS clinical data,survival and longitudinal data may be the best way to reduce such bias and improve model accuracy(Diggle2009,Dimitris2012,Elashoff2016,etc.)However,in the theoretical and applied research of joint model construction,the choice of variables has been a difficult problem that can not be solved well(Wu 2012,Elashoff2016).Since 1999,the variables based on punishing least squares or punishing speculative thoughts(Such as adaptive LASSO method,etc.)have been developed rapidly and have excellent Oracle properties in the variable selection calculation of Cox class model and mixed effect class model(Fan 2001,Fan 2002,Fan 2004,Zou2006,Fan2014,etc.)In this study,we try to apply the penalty likelihood method(Fan 2001,Fan 2002,Fan 2004,Zou2006)based on SC any implicit penalty deviation function to the variables of surviving part and longitudinal part model in joint model construction The model of the two sets of real clinical data is modeled and compared with the results of previous analysis.The advantages and disadvantages of the model are analyzed and analyzed,and the scientific research basis and reference are provided for the clinical data research.Finally,we also explored a variable selection method based on adaptive LASSO proposed by He.in 2015.