proc surveyselect question: random sample by variable values?

[RobertWF1]

Say I want to select a random sample of 5 car makes (out of the total of 38 different makes) from sashelp.cars and pull all records sharing one of the randomly select 5 car makes.

I've tried the following but it returns an error message. 

How do you perform random sampling of records by variable value?

proc sort data=sashelp.cars; by make; run;

proc surveyselect data=sashelp.cars
out=work.cars_samp
seed=123
method=srs
n=5;
strata make;
run;

ERROR: The sample size, 5, is greater than the number of sampling units, 1.
NOTE: The above message was for the following stratum:
Make=Hummer.
ERROR: The sample size, 5, is greater than the number of sampling units, 2.
NOTE: The above message was for the following stratum:
Make=Isuzu.
ERROR: The sample size, 5, is greater than the number of sampling units, 3.
NOTE: The above message was for the following stratum:
Make=Jeep.
ERROR: The sample size, 5, is greater than the number of sampling units, 3.
NOTE: The above message was for the following stratum:
Make=Land Rover.
ERROR: The sample size, 5, is greater than the number of sampling units, 2.
NOTE: The above message was for the following stratum:
Make=MINI.
ERROR: The sample size, 5, is greater than the number of sampling units, 3.
NOTE: The above message was for the following stratum:
Make=Oldsmobile.
ERROR: The sample size, 5, is greater than the number of sampling units, 2.
NOTE: The above message was for the following stratum:
Make=Scion.
NOTE: The SAS System stopped processing this step because of errors.
WARNING: The data set WORK.CARS_SAMP may be incomplete. When this step was stopped there were 155 observations and 17 variables.
WARNING: Data set WORK.CARS_SAMP was not replaced because this step was stopped.

[SAS_Rob · Accepted solution]

Putting MAKE on the CLUSTER statement will have it select 5 different makes, keeping all the observations in each make.

proc surveyselect data=sashelp.cars
out=work.cars_samp
seed=123
method=srs
n=5;
cluster make;
run;
proc freq data=work.cars_samp;
tables make;
run;

[PGStats]

Select the makes, then extract the sample:

proc sql;
create table makes as
select unique make from sashelp.cars; 
quit;
 
proc surveyselect data=makes n=5 out=makes_5 seed=123; run;

proc sql;
create table cars_samp as
select *
from sashelp.cars
where make in (select make from makes_5);
drop table makes_5;
quit;