topic Re: Subtracting withing cluster. Unequal space and follow up Longitudinal data in SAS/IML Software and Matrix Computations

Subtracting withing cluster. Unequal space and follow up Longitudinal data

desireatem — Tue, 16 Sep 2014 17:18:09 GMT

Hello, I am trying to compute S using the data with variable N and time. See example below. My code below does not work:

N time S

1 2 2 S[1]=time

1 5 5-2 = time[i+1]-time

1 7 7-5 = time[i+1]-time

2 5 5 S[1]=time

2 7 7-5 = time[i+1]-time

2 8 8-7 = time[i+1]-time

2 10 10-8 = time[i+1]-time

3 1 1

3 7 7-1

3 9 9-7

3 9 9-9

proc iml;

use simulate1_&u;

read all var{N time } into DM;

close;

time = DM[,2]; N = DM[,1];

n = nrow(DM);

TT=J(nrow(time),1,1);

do i = 1 to n;

if i=1 then S=time;

else if 1 < i < =n then S=time[i+1]-time;

end;

create kaplann_&u var{time N S };

append;

quit;

Re: Subtracting withing cluster. Unequal space and follow up Longitudinal data

Rick_SAS — Tue, 16 Sep 2014 18:55:21 GMT

There are several ways to do this. If the N variable is unsorted, you could use the UNIQUE-LOC technique. I assume the data are sorted by N, in which case it is more efficient to use the UNIQUEBY technique.

There is no need to loop over the observations to compute time-time[i-1]. Instead you can use the DIF function.

data a;
input N time @@;
datalines;
1 2 1 5 1 7 2 5 2 7 2 8 2 10 3 1 3 7 3 9 3 9
;

proc iml;
use A;
read all var{N time}; /* assume data sorted by N */
close;

nobs = nrow(time);
S = J(nobs,1,.);
b = uniqueby(N, 1); /* b = beginning of i_th category */
b = b // (nrow(N)+1); /* trick: append (n+1) to end of b */

do i = 1 to nrow(b)-1;
   idx = b:(b[i+1]-1);    /* Find observations in i_th level */
   S[idx] = dif(time[idx]); /* DIF of time steps */
   S[idx[1]] = time[idx[1]]; /* replace missing with 1st time */
end;

print S;

Re: Subtracting withing cluster. Unequal space and follow up Longitudinal data

desireatem — Tue, 16 Sep 2014 19:06:58 GMT

My data is simulated data, so it not fixed. That was just an example. I think I can fix your code to work with my simulation.

Re: Subtracting withing cluster. Unequal space and follow up Longitudinal data

Rick_SAS — Tue, 16 Sep 2014 19:11:16 GMT

If you read the articles, they describe how to handle other cases, such as unsorted data.

Re: Subtracting withing cluster. Unequal space and follow up Longitudinal data

desireatem — Wed, 17 Sep 2014 00:09:52 GMT

I tried to modify your code to to both sum and diffrence but it does not work:

proc iml;

use simulate1_&u;

read all var{TIME OLF SURVIVAL1} into DM;

close;

SURVIVAL1 = DM[,3]; OLF = DM[,2]; TIME = DM[,1];

n = nrow(DM);

segment = J(nrow(DM),1,1); *print segment;

SS=J(nrow(time),1,1);

nobs = nrow(time);

Survival1 = J(nobs,1,.);

a = uniqueby(time , 1); /* a = beginning of i_th category */

a = a // (nrow(time )+1); /* trick: append (n+1) to end of a */

do i = 1 to nrow(a)-1;

idx = a:(a[i+1]-1); /* Find observations in i_th level */

SS[idx] = sum(survival1[idx]); /* sum of time steps */

SS[idx[1]] = survival1[idx[1]]; /* replace missing with 1st survival1 */

TT=J(nrow(survival1),1,1);

nobs = nrow(survival1);

Survival1 = J(nobs,1,.);

b = uniqueby(Survival1 , 1); /* b = beginning of i_th category */

b = b // (nrow(Survival1 )+1); /* trick: append (n+1) to end of b */

do i = 1 to nrow(b)-1;

idx = b:(b[i+1]-1); /* Find observations in i_th level */

TT[idx] = dif(time[idx]); /* DIF of time steps */

TT[idx[1]] = time[idx[1]]; /* replace missing with 1st time */

end;

create kaplann_&u var{time OLF survival1 SS TT };

append;

quit;