topic Random Sample taken from a balanced panel in Statistical Procedures

Random Sample taken from a balanced panel

GKati — Sat, 26 Aug 2017 20:14:39 GMT

Hello,

I have a large balanced panel of clients (each client is present in each period). I would like to take a random sample by randomly picking the client and keeping information for all years for that client.

What's the easiest way to do this?

Re: Random Sample taken from a balanced panel

PGStats — Sat, 26 Aug 2017 21:22:07 GMT

1) Create a list of all the clients (proc SQL)

2) select a sample of clients (proc surveyselect)

3) extract all data for those sample clients (proc SQL)

Re: Random Sample taken from a balanced panel

GKati — Sat, 26 Aug 2017 21:28:49 GMT

How exactly do I do 3. in SQL? I can just extract based on a list of client numbers in another database?

Re: Random Sample taken from a balanced panel

PGStats — Sat, 26 Aug 2017 22:05:59 GMT

Assuming that step 2 results in a table called sampleClients, do something like :

proc sql;
create table samplePanel as
select * 
from myPanelData
where clientId in (select clientID from sampleClients);
quit;

Re: Random Sample taken from a balanced panel

PGStats — Sun, 27 Aug 2017 02:56:07 GMT

Actually, there is a simpler way to do this using surveyselect. The trick is to do cluster sampling using the client as the cluster:

data panel;
do clientID = 1 to 20;
    do period = 1 to 10;
        value + 1;
        output;
        end;
    end;
run;

proc surveyselect data=panel out=panelSample sampsize=10 seed=868585;
samplingunit clientID;
run;