[15.068 - Statistical Consulting] Generating a random sample - using state lotteries!
How do you pick a random sample of 300 transactions from a population of 60,000 credit card transactions?
Key is - how do you convince people that the process used is truly unbiased and random? Say people are suspicious and you have to come up with something that will stump them completely.
Idea - Rank the transactions some way - say you rank them chronologically with a procedure for breaking a tie (e.g. alphabetic by last name).
Idea - use a "public source of random numbers", like the state lottery. If the lottery numbers don't match correctly, just concatenate and drop numbers to make a very easy algorithm to get the exact number of digits needed.
Use New York state lottery if client is in New York :-)
Idea - Let the client pick the starting date when you start looking up the lottery numbers.
What's good? Easy to understand, easy to replicate.
2:10:04 PM
|