fdarnell
Well-known
It would be the z test for a proportion where we test the hypothesis that p is less or equal to 50% versus p exceeds 50%. With such a huge "sample", of course the test will be significant. The difference between BREXT support and REMAIN support is non-zero, and it favored BREXT.
This is a statistically significant difference, but is it a meaningful difference on the ground? In clinical trials, we often hear a similar argument. It is of statistical significance, but is it of clinical significance? Drug X shows a 0.1% improvement in the patients over Drug Y. Is it of clinical relevance?
My statistics is limited to photography, unfortunately. We would only say the film is good, or no-good. This held until an old Japanese gentleman told me there was another case, no-good, but ok!. Perhaps the clinical side defines these kind of cases.
I am lucky I no longer need stats to earn an income, mine is quite dated. Worked for that old plus-x though!😉