Date: Dec 25, 2012 2:55 PM
Author: Luis A. Afonso
Subject: Almost surely one is able to detect differences . . .
Not surprisingly there are a lot of persons rather uncomfortable with the random event?s nature . . . The anti-NHST people has a handful arguments that unfortunately has some impact on insufficiently learned ones.
Here one of them was chosen: What is the usefulness of a tool that, gathering a sufficiently data, we are always leading to reject a Null Hypotheses? Not to reject is surely synonymous of enough data collected.
They are absolutely right. . . except on a `mince´ detail: the usefulness.
What could be counter claimed?
___We are before the result of a concrete experiment which is depending on all conditions conditioning it. Sample size for a physicist is like a resolution power: if small or large it does disallow to detect little things or allows to. I do not feel surprised if a small sized sample admits/not rejects almost whatever hypotheses; and likely if very large ones are able to see no mater small differences. Anti-NHST people seem it was demanded that a difference (of parameters) must be noticed whatever the number of items sampled. Foolish.
I can state that a sample can carry a lot of information about its parameters if sufficiently large or contrarily do not if small or not random.
Luis A. Afonso