Drexel dragonThe Math ForumDonate to the Math Forum

Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Math Forum » Discussions » sci.math.* » sci.stat.math

Topic: Who cares the null hypotheses (two-tail test) is really TRUE?
Replies: 2   Last Post: Dec 22, 2012 7:42 PM

Advanced Search

Back to Topic List Back to Topic List Jump to Tree View Jump to Tree View   Messages: [ Previous | Next ]
Luis A. Afonso

Posts: 4,758
From: LIsbon (Portugal)
Registered: 2/16/05
Who cares the null hypotheses (two-tail test) is really TRUE?
Posted: Aug 5, 2011 6:10 PM
  Click to see the message monospaced in plain text Plain Text   Click to reply to this topic Reply

Who cares the null hypotheses (two-tail test) is really TRUE?

Not me, in accordance to learned people! DEFINITIVELLY! Seeming a blasphemy, namely for those that state: H0 shows not be true if a sufficiently large sample is analysed . . . because an even a tinny difference could be detected with substantial probability, for example 99.99%.
Nothing more WRONG! The mater is absolutely other: we cannot prove the Null Hypothesis true or not true . . . IT IS IMPOSSIBLE: the most can be stated is that we should reject the null, or, on contrary, it was found no sufficient evidence to do so. Not more, not less.
The null for a parameter as been associated to mu=0: NO, is preferable to try
____1)____H0: | mu - mu0 | <= h_______ or else
____2) Ha: | um - mu0| > h
If the test statistics imply 1) we say that the obtained mu is so close to the standard mu0 that is not greater than h positive (null *accepted*), otherwise, resulting 2), indicates that the difference is at least h, at the chosen significance level, alpha. This no statistical quantity is arbitrary chosen, depending the case on study, as the minimum *interesting to research* or *economically* that worth to be stated/published as SIGNIFICANT. People that, in this instance, do not agree to let h=0 is fully right: a mu value that is only statistically significant does not indicate really in what concerns how much is different from mu0, in spite of even, perhaps, showing a very small alpha. The *insignificance* of a (statistically) significant result: what one can say is only that is well apart mu0.

Luis A. Afonso

Tossing coin:

_ 1- alpha = 0.99____ Z= [ -2.807, 2.807]

____Z= |.505 - .5| / sqrt (.5*(1-.5)/n) =2.807
_______.005/2.807 = sqrt (.5*.5 / n)
________sqrt(n)= .5*2.807/.005 = 280.7
___________n = 78792.49 +/- 394 tosses

With a difference of +/- 0.005 we should perform 79792 tosses and to find heads (or tails) in [78398, 79186] in order not to reject 1% alpha the coin is fair.

Point your RSS reader here for a feed of the latest messages in this topic.

[Privacy Policy] [Terms of Use]

© The Math Forum at NCTM 1994-2016. All Rights Reserved.