topic Re: Does the order matter while resetting values in RETAIN statement in SAS Programming

Does the order matter while resetting values in RETAIN statement

kingCobra — Mon, 02 Nov 2020 12:28:50 GMT

Hi all,

I have had a theoretical question in mind for long time.

Does the order matter while resetting values in RETAIN statement?

Ex -

proc sort data=have;

by subjid parameter visit;

run;

data want;

retain total 0;

set have;

by subjid parameter visit;

if first.parameter then total=0;
if variable1='NOT DONE' then total=total+1;

run;

Does it produce the same result if I change the order of the statement marked in red?

That is

1. if variable1='NOT DONE' then total=total+1;

2. if first.parameter then total=0;

In some complex cases I think I have seen the resetting statement (if first.parameter then total=0;) at #1 does not actually reset the value of total. But I could not found any theoretical basis for the ordering of the statements.

Any help will be appreciated.

Re: Does the order matter while resetting values in RETAIN statement

PaigeMiller — Mon, 02 Nov 2020 12:32:14 GMT

You could try it and find out (and let us know what happens).

Re: Does the order matter while resetting values in RETAIN statement

RichardDeVen — Mon, 02 Nov 2020 13:03:23 GMT

The RETAIN statement is not executable, it is a compile time statement that sets the initial value of the variable.

Example:

In the sample the extra emphasis on compile time behavior is shown by placing the RETAIN statement at the bottom of the step. The first PUT shows the initialization value.

data _null_;
  put x=;
  set sashelp.class;
  retain x 123;
run;

For your case of REATAINing a total value that incrementally counts the number of NOT DONEs over a parameter within a subjid, the order of statements absolutely matters. The "total" might be better named or labeled "how_many_not_dones_so_far"

The first. conditional must occur first

if first.parameter then total=0;
if variable1='NOT DONE' then total=total+1;

Consider what would happen if the statements were reversed, AND, (edge case) the data had NOT DONE in the first row of the group?

if variable1='NOT DONE' then total=total+1;
if first.parameter then total=0;

In the (edge case) your total would be one less that what it should be because the total starting at first.parameter state should be 1, not 0.

Re: Does the order matter while resetting values in RETAIN statement

kingCobra — Mon, 02 Nov 2020 13:26:41 GMT

Thanks Richard for the detail explanation. I just constructed the example dataset and choose random names for the variables. My idea was to get the theoretical basis for RETAIN statement.
In my case, I was getting 1 more count as the previous subjid has variable1='NOT DONE' as the last record.