06-20-2017 11:41 AM
The lag() function works like a FIFO (first-in-first-out) queue. It only gets a value when it is called. Calling it conditionally will result in not all values being fed into the queue, and therefore unwanted results.
06-20-2017 02:31 PM
I leave that as an exercise for the reader. It's really simple to devise code that illustrates this.
@KurtBremser I wonder if you have an example for this?
06-22-2017 03:56 PM
Let's I have a mock data set.
When I execute lag(x), what would be the value in queue 3 or 1? Is there anyway we can see the values in queue?
06-22-2017 06:02 PM
so here is my understanding, when x = 1, there is nothing in the queue, hence it returns missing value. When X=2, 1 is in the queue and it returns 1. Similarly when x =3, 2 is in the queue, and it returns a value of 2.
06-22-2017 06:26 PM - edited 06-22-2017 06:28 PM
Your understanding is correct. I don't think there s a way to see what is inside the queue, other than pushing values into the queue to push them out.
Definitely take the time to play around with conditional lags, to convince yourself that the lag() function works when called conditionally. Far too many people mistakenly think there is a bug in the lag function and that it should never be called conditionally. It works perfectly fine conditionally, you just need to understand how a queue works. e.g., think about what you should expect from:
data _null_; set mock; if x IN (2,3) then lag=lag(x); put x= lag=; run;
Also worth playing with lags with longer queues, e.g.:
data _null_; set mock; lag1=lag1(x); lag2=lag2(x); lag3=lag3(x); put (x lag1 lag2 lag3)(=); run;
06-23-2017 02:25 AM
To slightly expand your code and show the problems of a conditional lag, just do this:
data mock; input x; cards; 1 2 3 4 5 ; run; data test; set mock; if x ne 3 then x1 = lag(x); run; proc print data=test noobs; run;
The result shows the effect of lag() not being called in obs 3:
x x1 1 . 2 1 3 . 4 2 5 4