I have the following situation:

  • 6,000 letters created in July
  • 250 staff created the letters
  • Some staff created more letters than others

I'm looking to do a quality assurance exercise where we check the quality of a sample of the letters created by the staff. I need some help to understand how we should prepare the sample for this one as there are two population types - 1 the number of letters created, and 2 the number of staff who created the letters.

Should we:
  1. Randomly sample from the 6,000 letters even though this may mean that some staff have 2-3 letters checked and other staff have 0, or
  2. Control the sample so that there is at least 1 letter from each staff member in the sample size (assuming the total sample size is around 350)?
What type of sampling is option 2 and would this be more or less statistically valid than simple random sampling?

Any help or guidance here would be great!