During a class exercise designed to demonstrate sampling distributions, I had each of my students use the Data>Select Cases>Random sample of cases procedure to randomly select a sample of records from a large data file containing nearly 3,000 observation. All students requested a random sample of 4% of the entire data file.

All of us ran descriptive statistics on our own randomly selected subset of the data. As we were discussing our sample means, we quickly realized that our averages were identical.

We double checked to make sure that we all chose the random selection option, not the "first ____ records of ____ " option. I can confirm that the argument to select a random percentage of all cases was set up correctly, but we did NOT get different random samples of our large date file.

Is this a known issue or is the random sample of cases procedure not the procedure it suggests it is?

We are using SPSS v26 in the cloud.

Thanks!

------------------------------
Timothy Hart, PhD
Department of Criminology and Criminal Justice
University of Tampa
Tampa, FL
------------------------------

#SPSSStatistics

Changing the random number seed worked! Thanks, Jon.

------------------------------
Timothy C. Hart
University of Tampa
Tampa FL
8132573329
------------------------------

Original Message:
Sent: Fri October 09, 2020 04:49 PM
From: Jon Peck
Subject: Selecting a Random Sample of Cases

Check your setting for the random number seed via Transform > Random Number Generators.  You probably all have the same fixed starting point for the generator.

Also, you should change the setting to use the Mersenne Twister.  It has better randomness properties, but the default is still the old generator due to compatibility issues.

--

Jon K Peck
jkpeck@gmail.com