Now going to your presented case: 2 out of 20000 in sample size complained about something. Is this sample size representative to say that this is not statistically valid? This statistically is insufficient to prove that the matter is not representative. The correct way to approach whether this is relevant or not would be to ask for 2000 of them (because the error margin of that sample is around 2p.p) If they would like to have blue flowers instead.
There are thousands of single or dual mention items that you would have to ask people about. Would it be better in terms of the construct validity? Yes Is it feasible? No
You are talking about validity in the context of a test. As the available feedback does not probe for anything in particular I wonder in what way the critereon of test validity is applicable to it at all. It applies when you are measuring something specific.
What we have can provide the basis for probing specific items.
I think we agree that a survey should follow the unstructured interview
Sorry for the misspellings. I’m not native
No problem, I am not a native english speaker either. In some ways my Latin is better
