OPTIONS NOCENTER FORMCHAR = "|----|+|---+=";
TITLE1 'Homework Number Five - Arthritis across 3 timepoints';

DATA ONE; INFILE 'C:\DATA\arthritb.dat';
INPUT ID ARTHRIT TXAURA TIME;

/* produce time contrasts with baseline as the reference cell */
TIMEA = 0; TIMEB = 0;
IF TIME EQ 1 THEN TIMEA = 1;
IF TIME EQ 2 THEN TIMEB = 1;

/* get observed proportions across time by treatment */
PROC FREQ; TABLES TIME * TXAURA * ARTHRIT ;
RUN;

/* convert data to multivariate setup */
DATA NEW (KEEP = ID TXAURA ARTHRIT1-ARTHRIT3);
     ARRAY A(3) arthrit1-arthrit3;
     DO I = 1 TO 3;
         SET ONE;   BY ID;
         A(I) = ARTHRIT;
         IF LAST.ID THEN RETURN;
     END;
RUN;

/* get observed correlations of the dependent variable */
PROC CORR; VAR ARTHRIT1-ARTHRIT3;
RUN;


/* GEE MODEL */
/* note: the type of working correlation needs to be specified */
PROC GENMOD descending DATA=ONE;
  CLASS ID TIME;
  MODEL ARTHRIT = TIMEA TIMEB TXAURA*TIMEA TXAURA*TIMEB / LINK=LOGIT DIST=BIN;
  REPEATED SUBJECT=ID / WITHIN=TIME CORRW TYPE=??????;
run;