Matlab Codes for Data Analysis in the Paper

'Semiparametric marginal regression analysis for dependent competing risks under an assumed copula'
By Yi-Hau Chen (yhchen@stat.sinica.edu.tw)



Three Matlab scripts are provided (also downloadable from http://www.stat.sinica.edu.tw/yhchen/download.htm):

(1) dependent censoring.m

    the main script for reading the data and calling for the 'fminunc' function of Matlab)
    model specifications of the method in the paper are specified in this script, including 
    (i) the copula model, specified by 
        
        fa='c';  for Clayton copula
        fa='f';  for Frank copula
        fa='g';  for Gumbel copula

    (ii) the covariate sets for the two marginal regressions, specified by, for example,

         w=[age, gender, treatment,race];
         w2=[age, gender,treatment];

    (iii)the marginal regression models for the two competing risks, specified in the way of 
        the generalized odds models in Dabrowska and Doksum (1988); for examples:
 
        r0=0;  
        r02=0;

        for both risks specified as the Cox proportional hazards models 

        r0=1;
        r02=0;

        for the first risk specified as the proportional odds model and
        the second risk specified as the proportional hazards model

    (iv) the Kendall's tau between the two risks, specified by,  for example,

        tau=0.3;


(2) ffff_d.m

    the script for calculating the score and information matrix of the likelihood


(3) ggg.m

    the script for calculating quantities required in calculating the likelihood function and the score