SAS not processing if then statement following first conditional set of statements??

[silversta]

Hello SAS community,

Everyone in my dataset has an outcome date (and some people have two outcome dates). In the code below (where the line starts with "if tv_outcome=1 then do..."), SAS is not processing the statements. For those people who have two outcomes, SAS processes the statements when tv_outcome=0 and tv_outcome=1, but then doesn't process the last set of statements that begin with "if tv_outcome=2 then do...". For example, there are people who have missing date_of_event and missing second_outcome_date, but instead of calculating time_to_event as obs_end-date_of_outcome, it just sets it to missing. So the data look something like this:

id    obs_beg          obs_end      tv_outcome   date_of_outcome   date_of_event    time_to_event  

1    14MAR2010     31OCT2015        0                12APR2012                 .                       760

1    14MAR2010     31OCT2015        1                12APR2012                 .                       .

This is how I need it to look:

id    obs_beg          obs_end      tv_outcome   date_of_outcome   date_of_event    time_to_event  

1    14MAR2010     31OCT2015        0                12APR2012                 .                       760

1    14MAR2010     31OCT2015        1                12APR2012                 .                       1297

Code:

data event_time (keep = id obs_end death_dt date_of_outcome date_of_event
second_outcome_date time_to_event tv_outcome event_ind);
set eventression;
format obs_end date9.;
obs_end = '30sep2015'd;
if death_dt ne . then do;
obs_end = death_dt;
end;
if tv_outcome=0 then do;
if date_of_event = . then time_to_event=date_of_outcome-obs_beg;
else if date_of_event ne . and date_of_event gt date_of_outcome then time_to_event=date_of_outcome-obs_beg;
else time_to_event=date_of_event-obs_beg;
event_ind=0;
end;
if tv_outcome=1 then do;
if date_of_event = . and second_outcome_date = . then time_to_event=obs_end-date_of_outcome;
else if date_of_event = . and second_outcome_date ne . then time_to_event = second_outcome_date-date_of_outcome;
if date_of_event ne . and date_of_event lt date_of_outcome then time_to_event=.;
else if date_of_event ne . and second_outcome_date ne . and date_of_event gt second_outcome_date then time_to_event=second_outcome_date-date_of_outcome;
else time_to_event=date_of_event-date_of_outcome;
event_ind=1;
end;
if tv_outcome=2 then do;
if date_of_event = . and second_outcome_date ne . then time_to_event=obs_end-second_outcome_date;
if date_of_event ne . and date_of_event lt second_outcome_date then time_to_event=.;
else time_to_event=date_of_event-second_outcome_date;
event_ind=2;
end;
*if time_to_event<0 then time_to_event=0; *this will be negative values;
run;

proc sort data=event_time nodupkey out=event_time2; by id tv_outcome;
run;

Any help would be most appreciated!

[Tom]

None of your example data have TV_OUTCOME=2.  Why would it execute those statements when none of your data match this pattern.

[silversta]

Hi Tom,

I provided a simple example to show that this doesn't execute where tv_outcome=1, I assumed I could figure it out for the other scenario once I get it to execute for the first set of statements. However, if you would like to see where the issue lies with tv_outcome=2, please see the example below. 

id    obs_beg          obs_end      tv_outcome   date_of_outcome   second_outcome_date   date_of_event    time_to_event  

2    14MAR2010     31OCT2015        0                12APR2012                24Mar2014                       .                       760

2    14MAR2010     31OCT2015        1                12APR2012                24Mar2014                       .                       .         

2    14MAR2010     31OCT2015        2               12APR2012                 24Mar2014                       .                       .            

This is what I need.

id    obs_beg          obs_end      tv_outcome   date_of_outcome   second_outcome_date   date_of_event    time_to_event  

2    14MAR2010     31OCT2015        0                12APR2012                24Mar2014                       .                       760

2    14MAR2010     31OCT2015        1                12APR2012                24Mar2014                       .                       711      

2    14MAR2010     31OCT2015        2               12APR2012                 24Mar2014                       .                       1297           

[PGStats]

Aren't you missing some else

data event_time (
	keep = id obs_end death_dt date_of_outcome date_of_event second_outcome_date time_to_event tv_outcome event_ind);
set eventression;
format obs_end date9.;
obs_end = '30sep2015'd;
if death_dt ne . then do;
	obs_end = death_dt;
	end;
if tv_outcome=0 then do;
	if date_of_event = . 	then 
		time_to_event=date_of_outcome-obs_beg;
	else if date_of_event ne . and date_of_event gt date_of_outcome  then 
		time_to_event=date_of_outcome-obs_beg;
	else 
		time_to_event=date_of_event-obs_beg;
	event_ind=0;
	end;
if tv_outcome=1 then do;
	if date_of_event = . and second_outcome_date = . then 
		time_to_event=obs_end-date_of_outcome;
	else if date_of_event = . and second_outcome_date ne . then 
		time_to_event = second_outcome_date-date_of_outcome;
	else if date_of_event ne . and date_of_event lt date_of_outcome then 
		time_to_event=.;
	else if date_of_event ne . and second_outcome_date ne . and date_of_event gt second_outcome_date then
		time_to_event=second_outcome_date-date_of_outcome;
	else 
		time_to_event=date_of_event-date_of_outcome;
	event_ind=1;
	end;
if tv_outcome=2 then do;
	if date_of_event = . and second_outcome_date ne . then 
		time_to_event=obs_end-second_outcome_date;
	else if date_of_event ne . and date_of_event lt second_outcome_date then 
		time_to_event=.;
	else 
		time_to_event=date_of_event-second_outcome_date;
	event_ind=2;
	end;
*if time_to_event<0 then time_to_event=0; *this will be negative values;
run;

Without those, your second IF statement overwrites the first.

[Kurt_Bremser]

With spaghetti code like that, it is no miracle at all that you lose track of your functional blocks. Apply some visual formatting and see if it gives you a clue.

data event_time (
  keep=
    id obs_end death_dt date_of_outcome date_of_event
    second_outcome_date time_to_event tv_outcome event_ind
);
set eventression;
format obs_end date9.;
obs_end = '30sep2015'd;
if death_dt ne .
then do;
  obs_end = death_dt;
end;
if tv_outcome = 0
then do;
  if date_of_event = .
  then time_to_event = date_of_outcome - obs_beg;
  else if date_of_event ne . and date_of_event gt date_of_outcome
  then time_to_event = date_of_outcome - obs_beg;
  else time_to_event = date_of_event - obs_beg;
  event_ind = 0;
end;
if tv_outcome = 1
then do;
  if date_of_event = . and second_outcome_date = .
  then time_to_event = obs_end - date_of_outcome;
  else if date_of_event = . and second_outcome_date ne .
  then time_to_event = second_outcome_date - date_of_outcome;
  if date_of_event ne . and date_of_event lt date_of_outcome
  then time_to_event = .;
  else if
    date_of_event ne . and second_outcome_date ne . and
    date_of_event gt second_outcome_date
  then time_to_event = second_outcome_date - date_of_outcome;
  else time_to_event = date_of_event - date_of_outcome;
  event_ind = 1;
end;
if tv_outcome = 2
then do;
  if date_of_event = . and second_outcome_date ne .
  then time_to_event = obs_end - second_outcome_date;
  if date_of_event ne . and date_of_event lt second_outcome_date
  then time_to_event = .;
  else time_to_event = date_of_event - second_outcome_date;
  event_ind = 2;
end;
*if time_to_event < 0 then time_to_event = 0; *this will be negative values;
run;