Date: Jun 1, 2014 4:10 PM
Author: Luis A. Afonso
Subject: Taking a larger sample size ...
Taking a larger sample size ...
Some people do argue that the Null Hypothesis Significance Tests are ´worthless´ because we always are able to attain significant results by simply joining more data. And the result is indisputably correct. They are right. Congratulations!
These fellows, owing their complete presumptuous nature; finally find out the absolute argument they needed to annihilate NHST. For them, really, do not accept the idea that increasing the ´resolution power´ we see things/differences that were invisible/undetected before. I use to call the issue the ´Case of Optical Microscopy against Virus´ that raged all XIX century.
Even for ´exactly simulated´ data, say Normal/Gaussian by the Box-Muller algorithm, we only with probability ZERO find out a sample mean exactly equal to the Population mean value it was chosen for simulations . . .
The reason of dispute is that there are people that are not ready to accept that the conclusions gotten by observation/experiment are contingent/dependent from the conditions under what we act. For them when H0 is false we must have rejection whatever the sample size. Is, not more, than the Classical Logic wrongly used in Random Phenomena area.
Luis A. Afonso