Unexpected results with lag function

[NickS2]

I am getting unusual results when I am using the lag function and not getting the results I am looking for. 

Dataset:

ID   Event_Date

1     12MAY2020

1      24JUN2020

1      08AUG2020

2      09APR2020

3      26JUN2020

3      30AUG2020

Program I wrote:

data want;

     set have;

     by ID;

     if first.ID then prev_event_dt = .;

     else prev_event_dt=lag(Event_Date);

run;

Output I am expecting:

ID   Event_Date               prev_event_dt 

1     12MAY2020              .

1      24JUN2020              12MAY2020

1      08AUG2020              24JUN2020

2      09APR2020              .

3      26JUN2020              .

3      30AUG2020             26JUN2020              

Please point out where I am going wrong with the program I wrote.

Thank you

[PaigeMiller]

LAG does not behave as you expect it in IF / THEN / ELSE statements. (It's doing the right thing, but your expectation is wrong)

Better you should use the IFN function (or IFC if you want to create a character variable)

prev_event_dt=ifn(first.id,.,lag(event_date));

[Tom · Accepted solution]

You never passed in the date from the first observation per ID into the queue/stack that LAG() uses. 

So it is not there to be returned later.

In general you never want to conditionally run the LAG() function because of this issue.

data want;
   set have;
   by ID;
   prev_event_dt=lag(Event_Date);
   if first.ID then prev_event_dt = .;
run;

[Kurt_Bremser]

The LAG function creates a FIFO queue that is filled every time the function is called. Calling LAG in a conditional branch means that not all values will make it into the queue.

Restructure your code:

data want;
set have;
by ID;
prev_event_dt = lag(Event_Date);
if first.ID then prev_event_dt = .;
run;