Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
NCTM or The Math Forum.


Luis A. Afonso
Posts:
4,758
From:
LIsbon (Portugal)
Registered:
2/16/05


Concerning Douglas H. Johnson and NHST
Posted:
Oct 19, 2013 11:27 AM


Concerning Douglas H. Johnson and NHST (Behavioral Ecology vol.16 no.1, 2005, http://beheco.oxfordjournals.org/content/16/1/323.full)
Strictly in Statistical terms we do intend to discuss the not unvaluable theme how a not insignificant number of users from other scientific branches negatively criticised the Null Hypotheses Statistical Tests´s worth. By decreasing order of seriousness one have:
___1___The null is near always false. (quoting)... most null hypothesis tested are known to be false before any data be collected . . .(end). Comment: NHST does not ever claims to be able to establish an hypothesis´s truthfulness, not at all, as it wellknown. On contrary since Ronald Fisher it was been well understood the only thing we can state is that there is (or not there is not) sufficient evidence the Null is false. The article A. induces otherwise.
___2___Zero as the parameter´s value (. . .) I argued that, instead of testing a hypothesis that the value of certain parameter is zero, it is often much more valuable to provide an estimate of that parameter, as well as is confidence interval (. . .) Comment: Reinforcing that he is full convinced that the null can be stated true he does not hesitate in ascribe a value to a parameter as the test aim, instead to say the advisable: that there is no sufficient evidence the parameter is zero. The confidence interval related with H0, yes, is somewhat enlightening because show the parameters variability bounds. ___3___The pvalue (. . .) Colegrave and Ruxton (2003) ascribes to Johnson: ?the p value is the probability that the null is actually true given the data? what is promptly denied. What he said is that to think so is a fantasy. In fact it is. There is no means to calculate from a sample that probability: the inductiveness the random events are fatally restrained forbids us to such goal. Statisticians uses a rather broader terminology: low pvalues if observed (at a rigorous test) is a symptom that H0 is unlike to be true.
___3___C.I. Breath or Location?
(. . .) The breath . . . gives an indication of the likelihood of the real effect size being zero (or very small) . . . (Colegrave and Ruxton) to which D.H.J. Said it is wrong preferring the C.I. location. Both are right: the later because the test statistics should be inside zero, the former is right in arguing that the narrow the interval the higher the precision. See my Sci Stat math Independent post :NHST : Accuracy and Precision
Well . . .
Luis A. Afonso



