3 ms·
Can someone explain me how Theranos' claim is anything else than me claiming that I can do a representative study with 5 participants and have the same statisti
by _pmf_ 10y ago
Can someone explain me how Theranos' claim is anything else than me claiming that I can do a representative study with 5 participants and have the same statistical significance as others doing a study with 10000 participants?
Isn't this just elementary statistics and completely independent from the domain?
- DougBTX 10y agoIf it is known beforehand that all the participants will answer in the same way, testing 5 will produce as much information as testing 10000, so the question is whether any beforehand assertion like that applies. One reason that assertion might not hold is if the 5 being tested come from a special interest group, knowing if that could be the case or not depends on domain knowledge. http://www.bd.com/vacutainer/labnotes/Volume17Number1/composition_differences.asp http://www.bd.com/vacutainer/labnotes/Volume17Number1/compos... So, it is known that small samples of blood taken from capillaries will have a different composition than large samples taken from veins. Is it possible to write a test that compensates for the known differences? Or is it just known that they are so different that one can't be inferred from the other? Hard to tell, needs cutting edge domain knowledge.
- umanwizard 10y agoYou're confusing random variation (which is helped by sample size) and bias (which isn't)
- DougBTX 10y agoMy example in the first paragraph has no random variation to explain how a test of 5 people could provide the same information as a test of 10000. The second example includes bias as a lead-in to the actual problems that tests of capillary blood need to handle.