Longitudinal data

[art297]

: FWIW, I wouldn't recommend using the monotonic() function as it could easily introduce a wrong answer.  There MUST be a reason why it has remained undocumented all of these years.  Personally, I wouldn't take the risk.

[Haikuo]

To Astounding: I hate this mouthful of hash() approach as well, the mere intention of it is to construct a one-pass solution, which in the end is proven to be lower efficient by Art. So that doesn't work out.

To Art and Fareeza: Agreed with your comments on last record. It only takes minor tweak to the current Hash() code to fix that:

data have;

input (fn ln) (:$20.) grade;

cards;

Ross Smith 11

Jayden Woo 10

Ross Smith 7

Jayden Woo 8

Ross Smith 6

;

/*theorically one-pass(I/O) approach*/

data want;

  declare hash full(ordered:'y');

  full.definekey('_n_');

full.definedata('fn','ln','grade');

full.definedone();

declare hiter hif('full');

declare hash hoh();

hoh.definekey('fn','ln');

hoh.definedata('fn','ln','lst','hil');

hoh.definedone();

  declare hash lst;

declare hiter hil;

do _n_=1 by 1 until (done);

  set have end=done;

l_grade=grade;

  if hoh.find() ne 0 then do;

lst=_new_ hash (ordered:'d'); 

lst.definekey('_n_');

lst.definedata('fn','ln','l_grade');

lst.definedone();

  hil=_new_ hiter ('lst');

hoh.replace();

end;

  lst.replace();

full.replace();

end;

do _rc=hif.next() by 0 while (_rc=0);

hoh.find();

hil.first();

output;

_rc=hif.next();

end;

stop;

drop _rc;

run;

Haikuo