What is the difference between survey errors in coverage vs. sampling?

Question

I'm taking an online course in Usability Surveys.

Both of these seem to be about missing some group from the sample. The example in the course is: You are surveying people about a new HR Form. You leave out the janitors. That's a sampling error. You email the survey but folks without email don't get it, that's a Coverage Error.

But another definition I found suggests that sampling error is more like "the sample you surveyed wasn't large enough" (statistically valid).

Which definitions should I use for coverage and sampling errors?

Answer 1

Whether "leave out the janitors" counts as sampling or coverage error depends on why they were left out.

If you randomly sampled from a staff list which included the janitors, but by chance your random number generator didn't pick any janitors, that would be sampling error. If you increased the sample size - or did the survey again with a different bunch of random numbers - this error would probably change.

If your sampling method makes it impossible to select janitors, e.g. you're using a list of permanent employees but the janitors are on contract, that's coverage error. No matter how much you increase the sample size or repeat the survey, you'll never select any janitors, so this error won't change.

Working from a frame that doesn't include the janitors is an error that affects sampling, but it's not "sampling error" as statisticians use that term.