topic Re: Replace missing variables (na) by random numbers within range in SAS Programming

Replace missing variables (na) by random numbers within range

Cruise — Tue, 15 May 2018 18:30:44 GMT

How to replace missing variables (na) by random numbers between 1 through 10? numbers can be repeated.

data x;
input a $;
cards;
1
2
na
na
4
na
na
;

Thanks millions in advance.

Re: Replace missing variables (na) by random numbers within range

Rick_SAS — Tue, 15 May 2018 18:39:43 GMT

Shouldn't 'a' be a numeric variable?

Anyway, see the article "How to generate random integers in SAS".

If you are reading a numeric variable x, then the modern code is

if missing(x) then x = rand("Integer", 1, 10);

Here I am assuming that you want a random integer. If you are using an old version of SAS, see the article for an alternative syntax.

Re: Replace missing variables (na) by random numbers within range

Cruise — Tue, 15 May 2018 18:41:07 GMT

Hi Risk, yes, it can be numeric. Thanks!

Re: Replace missing variables (na) by random numbers within range

Cruise — Tue, 15 May 2018 19:24:49 GMT

@Rick_SAS

Do you see anything wrong here? Month_of_birth is now a numeric variable with missing .

data want; set have;
if missing(Month_of_birth) then Month_of_birth=rand("Integer", 1, 31);
run;

ERROR: The first argument of the RAND function must be a character string with a value of
       BERNOULLI, BETA, BINOMIAL, CAUCHY, CHISQUARE, ERLANG, EXPONENTIAL, F, GAMMA, GAUSSIAN,
       GEOMETRIC, HYPERGEOMETRIC, LOGNORMAL, NEGB, NORMAL, POISSON, T, TABLE, TRIANGULAR, UNIFORM,
       or WEIBULL.

Re: Replace missing variables (na) by random numbers within range

Rick_SAS — Tue, 15 May 2018 19:30:09 GMT

As I indicated in my answer (and say explicitly in the article that I asked you to read), the "Integer" distribution was introduced in the past few releases of SAS. The error says that you are not using a recent version of SAS. See the article for how to define the %RandBetween macro and use %RandBetween(1,31).

Re: Replace missing variables (na) by random numbers within range

Cruise — Tue, 15 May 2018 19:43:18 GMT

Updating what worked:

%macro RandBetween(min, max);
   (&min + floor((1+&max-&min)*rand("uniform")))
%mend;

data success; set have; 
if Month_of_birth=. then Month_of_birth = %RandBetween(1, 12);
run;

Thanks @Rick_SAS