########## Oscars data , 20 selected male perfromers (who won or were nominated)

Data from SAS (Sas code for splitting records given at bottom of this file)

BIRTH    AGE_NOM1    AGE_WIN1    AGE_LAST    DEAD

 1971        8           .          30         0
 1962       31           .          39         0
 1961       34           .          40         0
 1930       36           .          50         1
 1888       52           .          53         1
 1899       44          52          58         1
 1942       34           .          59         0
 1901       35          40          60         1
 1929       38           .          60         1
 1941       56           .          60         0
 1939       23           .          62         0
 1936       61           .          65         0
 1935       53           .          66         0
 1933       56           .          68         0
 1916       44           .          72         1
 1923       39           .          73         1
 1925       33          61          76         0
 1919       46          46          77         1
 1888       42           .          84         1
 1911       33           .          90         0

# ph analysis in R, using "t.from, t.to" format for staggered entry

require(survival)

# this next analysis is INCORRECT.. it treats winner as a fixed variate,
# as though it were determined from the outset, when in fact it is dynamic..
# all start out as non-winners..

age.from = c( 8,31,34,36,52,44,34,35,38,56,23,61,53,56,44,39,33,46,42,33);
age.to   = c(30,39,40,50,53,58,59,60,60,60,62,65,66,68,72,73,76,77,84,90);
dead     = c( 0, 0, 0, 1, 1, 1, 0, 1, 1, 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 0);
winner   = c( 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0);

year.birth=c(1971,1962,1961,1930,1888,1899,1942,1901,1929,1941,
             1939,1936,1935,1933,1916,1923,1925,1919,1888,1911);

summary( coxph(Surv(age.from,age.to,dead) ~ winner) )


# BUT, by 'splitting' the records for the 4 who won, it be made into a form
# that allows "winner" to be treated as a dynamic variate.

age.from = c( 8,31,34,36,52, 44,52, 34, etc);
age.to   = c(30,39,40,50,53, 52,58, 59, etc);
dead     = c( 0, 0, 0, 1, 1,  0, 1,  0, etc);
winner   = c( 0, 0, 0, 0, 0,  0, 1,  0, etc);

# NOW  in

   coxph(Surv(age.from,age.to,dead) ~ winner) 
   
# "winner" be have the correct value in each riskset 
 


======== SAS split records, using the BIRTH  AGE_NOM1  AGE_WIN1 AGE_LAST DEAD

======== (above) as the source file ( "for634" )

data split; keep identity t_from t_to d w ;
 
 set for634;

 if age_win1 = . then do;
   t_from=age_nom1; t_to = age_last; d = dead; w = 0; output;
 end;

 if age_win1 ne . then do;
  if age_nom1 < age_win1 then do; t_from=age_nom1; t_to = age_win1; d = 0   ; w=0;output; end;
                                  t_from=age_win1; t_to = age_last; d = dead; w=1;output;
 end;
proc print; 

proc phreg data=split;

  model (t_from,t_to)*d(0) = w / ties=efron; 
  
Run;