defining new models for GOF testing in RDSURVIV

questions concerning analysis/theory using software from the Patuxent Software Archive

defining new models for GOF testing in RDSURVIV

Postby fanny » Wed Dec 08, 2004 5:08 am

hi!
i should test GOF values for several models on a set of data. Sadly enough, some of the models i'd have to test do not appear in the RDSURVIV output.

Is it possible to define a new model to be tested in RDSURVIV?
if so, how? a parameter in the command line?

(the model i should test is P(T,T),S(T),G(T),G2(T) )


thanks in advance
fanny
fanny
 
Posts: 11
Joined: Tue Nov 23, 2004 7:16 am

Postby jhines » Thu Dec 09, 2004 11:41 am

New models for RDSURVIV are created by writing model specifications with 'constraints'. The easiest way is to copy the text for a similar model, then add/change/delete constraint statements and give it a new name. For example to create a model with time-varying survival (Phi), copy the lines in the input file starting with 'PROC ESTIMATE' for the time-constant survival model, until the next 'PROC ESTIMATE' line, and delete the constraints dealing with PHI.

In the example on the RDSURVIV web-page, I would copy the following lines:

PROC ESTIMATE MAXFN=32000 NOVAR NAME=M(t)Stng;
INITIAL; RETAIN=A; CONSTRAINTS;
S(3)=S(2) /* c(1,2)=c(1,2) */;
S(6)=S(5) /* c(2,2)=c(2,2) */;
S(9)=S(8) /* c(3,2)=c(3,2) */;
S(12)=S(11) /* c(4,2)=c(4,2) */;
S(15)=S(14) /* c(5,2)=c(5,2) */;
S(18)=S(17) /* c(6,2)=c(6,2) */;
S(21)=S(20) /* c(7,2)=c(7,2) */;
S(28)=0 /* GAM'(2)=0 -> no emigration */;
S(29)=0 /* GAM'(3)=0 -> no emigration */;
S(30)=0 /* GAM'(4)=0 -> no emigration */;
S(31)=0 /* GAM'(5)=0 -> no emigration */;
S(32)=0 /* GAM'(6)=0 -> no emigration */;
S(33)=0 /* GAM'(7)=0 -> no emigration */;
S(34)=0 /* GAM(3)=0 -> no emigration */;
S(35)=0 /* GAM(4)=0 -> no emigration */;
S(36)=0 /* GAM(5)=0 -> no emigration */;
S(37)=0 /* GAM(6)=0 -> no emigration */;
S(38)=0 /* GAM(7)=0 -> no emigration */;
S(39)=1 /* THETA(1) not used in this model */;
S(40)=1 /* THETA(2) not used in this model */;
S(41)=1 /* THETA(3) not used in this model */;
S(42)=1 /* THETA(4) not used in this model */;
S(43)=1 /* THETA(5) not used in this model */;
S(44)=1 /* THETA(6) not used in this model */;
S(45)=1 /* THETA(7) not used in this model */;

and delete the lines containing "GAM" or "GAM'" to get a Phi(t),Gam(t),Gam'(t),P(t,t) model.
jhines
 
Posts: 502
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA


Return to analysis help

Who is online

Users browsing this forum: No registered users and 1 guest