proc surveyselect

[Jens_v]

Hallo SAS community.

I am stuck with a problem i hope you can help me with.

I have created the code:

proc surveyselect data= xxx.X201202  method=srs n=200 seed=200

      out=test1;

    Where department ne "AgroTech A/S";

Run;

However, i want to create a macro which loop through the data set

• Samlple1 .... Sample6

If the variable "Initials" from the sample1-6 is the same as initials in xxx.X201202 then these observationses should not be a part of the random sample of 200.

I hope you can help me.

/Jens

[art297]

I don't understand your question. If you simply want to randomly assign people to six different groups, without replication, then take a look at: 36383 - Randomly assign the observations in a data set to two or more groups

[Jens_v]

I have tried to make following code:

However, I have no idea if this is the right way to go. I hope you understand what my problem is...

[art297]

You have to indicate what you are trying to do. You're macro creates a local macro variable six times, each time discarding the previous version. I don't understand its purpose.

Your 2nd where statement doesn't make sense (to me). While you have shown any example data, a statement like:

where initialer ne 1, 3, 5;

won't work

I suggest you post some sample data, explain what you want to accomplish and, based on the sample data, what resulting file you want to obtain.

[Jens_v]

Hallo again and thank you for your patience.

Data set 1

Obs	Initials	Company	Cost
1	KLO	AgroTech A/S	50000
2	AVG	LEGO	12000
3	SLO	B&O	53000
4	NRL	AgroTech A/S	98000
5	PLA	INFINION	15244
n

Data set 2

Obs	Initials
1	AVG
2	KLP
3	QGY
4	DGY
5	MNJ
n

Data set 3

Obs	Initials
1	PLA
2	GTX
3	MLO
4	BBN
n	DYU

What I am trying to do is to make an random sample of "data set 1".

However, I do not want to include observations with the same initials from data set 2 and 3. Therefore the random sample must not include obs 2 and 5, because their initials occur in data set 2 and 3.

[art297]

Sounds like you are trying to do something like the following:

data initials;

  set sample1 sample2 sample3 sample4 sample5 sample6;

run;

proc sql noprint;

  select distinct quote(strip(initials))

    into :initials separated by ','

      from initials

  ;

quit;

proc surveyselect data=x201202 method=srs n=200 seed=200

  out=test005;

  where afdeling ne "AgroTech A/S" and

        initialer not in (&initials.);

run;

[Rick_SAS]

You might want to read the article "Sample with replacement in SAS" which discusses some of the issues and techniques for sampling with replacement.

[PGStats]

So, what I understand is that you already have samples and you want the new sample to exclude cases already selected. Try this (not tested) :

/* Create a  set of existing samples */

data sampSet / view=sampSet;

set xxx.sample:;

keep initials;

run;

/* Create a subset of x201202 excluding existing samples */

proc sql;

create view newSet as

select *

from xxx.X201202

where

  initials not in (select initials from sampSet);

quit;

/* Create a new sample from X201202, excluding previous samples */

call surveyselect data=newSet method=srs n=200 seed=200 out=test005;

run;

PG