keep first N observations for each factor-level combination

[blmreil]

Hi. I'm new to SAS so I'd greatly appreciate any help anyone could provide me. So say, I have a data set with ~500 observations and my ultimate goal is to run ANCOVA of this data set with response variable y, factors i and j, and possible covariates k and l. I have 5 levels of i and two levels of j. Since, I first want to try to run a completely balanced ANCOVA, I don't want to use the entire data set, just the first n observations (according to ID) for each of the ten factor-level combinations.

I initially wrote a line with

IF (ID > 10*n) then delete;

before realizing how stupid that is.

How would I go about doing this?

Here's a mockup of what part of the data looks like:

ID   i    j   k   l   y
1    3   1
2    3   1
3    2   1
4    4   1
5    5   2
6    1   1
7    4   2
8    2   2
9    3   1
10   5   2
11  1   1
12   3   2
13   2   1
14   1  2
15   5  1
...

So i=1, j=1, would have ID 11 as the first observation.
and i=2, j=1, would have ID 3, ID 13, etc.

It feels like this should be a pretty simple thing to do but I cannot for the life of me figure out how to do it. Would anybody be able to point me in the right direction? Thanks in advance for your help!

[KachiM]

Hi,

There are many ways to get the first N obsevations.

Use the statement as:

IF (ID > n) then delete;

 If your data set is named as MYDATA, then use;

set MYDATA(obs=n);

[blmreil]

Hi,

Thanks for your reply.  And sorry for not explaining it more clearly. I want to make it so each i-j combination has the same number of observations. If I do ID>200 or MYDATA(obs=200), it looks like it will just pull whatever the first 200 observations are. I want each combination to have 20 observations. 

[KachiM]

Yes. Now I see your requirement. It takes more time to give you solution without knowing some sample data set. Again you may want random sample in each i-j th combination. I have very limited time to give you a solution immediately. Someone will help you soon.

[KachiM]

Refer to the following and see Example 4:

https://stats.idre.ucla.edu/sas/faq/how-can-i-take-a-stratified-random-sample-of-my-data/