Pretpostavljam, da je vrlo malo ljudi procitalo tekst
"Statisticko zakljucivanje" na mojim web-stranicama, kao i moj clanak "Statistical 'Discoveries' and
Effect-Size Estimation", koji je objavljen u JASA 1989. godine.
( Journal of the American
Statistical Association, Vol. 84, No.
406 (Theory and Methods), 1989 (str.
608-610.) - Soric B.: "Statistical
'Discoveries' and Effect-Size Estimation" Taj moj clanak iz JASA ne objavljujem
na Internetu zbog copyrighta JASA. Separat
se moze dobiti od mene (jasno: besplatno). Inace, separati iz Journal of the American
Statistical Association (JASA) mogu
se naruciti na Internetu, ali se u tom slucaju placa desetak US dolara.
Web-adrese (JASA i formular za narudzbu):
Izgleda da jos uvijek ima malo onih istrazivaca koji shvacaju da
je vazno, u pojedinacnom istrazivanju (eksperimentu), postici sto visu razinu statisticke znacajnosti (najbolje p < 0.000001 = 10-6 ili,
jos bolje, p < 10-9).
Jos uvijek se smatra da se pojedinacna nul-hipoteza
moze odbaciti pri postignutoj razini znacajnosti od 5 posto (t.j. p = 0,05). Evo nekoliko primjera sa Interneta (nadjenih 23. VI. 2003.):
The Animated Software Company - Statstics Explained - Statistical Significance
"In Psychology, and in many other domains, it is customary to describe one's finding as statistically significant,
when the obtained result is among those that (theoretically) would occur no more than 5 out of 100 times (......) when (......)
random samples are drawn". - Last modified February, 2002 - Copyright (c) Russell D. Hoffman
Statistics Explained - Written by: Howard
S. Hoffman, Professor Emeritus of Psychology, Bryn Mawr College
"For most psychologists, and for many other scientists, it is customary to
set alpha at 0.05. (......) ....you are asserting that the odds of obtaining
that statistic by chance only are sufficiently low (one out of twenty) that it reasonable to conclude that your results are
not due to chance. Could you be in error? Of course you could, but at least you know the probability of such an error. It
is exactly equal to the value you have previously established for alpha". - Last modified February, 2002 - Webmaster: Russell D. Hoffman - Copyright (c) Russell D. Hoffman
Statistical Hypothesis Tests
William A. Levinson, P.E., MBA
"If we run an experiment whose result follows
a chi square distribution with 6 degrees of freedom....(.....) If chi square > 12.59, there is only a 5 percent chance
that it's just luck or variation, and we can be 95 percent sure that the null hypothesis is false".
Hyperstat Online Contents
"When using the 5% significance level, one concludes
that the experimental treatment has a real effect if chance alone would produce a difference as large or larger than the one
obtained only 5% of the time or less.
"The question now becomes, what
does it mean to be "statistically significant"? The answer is that you have a statistically significant effect if it is very
likely that H0 is false, meaning H1 is very likely to be true (......). The usual
values used for significance level are 0.05 or 0.01, and a result is "significant at the 0.05 level" if p(H0) < 0.05 (p(H0) is the probability H0 is true). Since
H0 is unlikely to be true, H1 is very likely to be true, and we accept the outcome of the experiment as statistically significant".
Neka manje-vise suprotna
Wikipedia, the free encyclopedia
"The value of the null hypothesis is that it can be rejected
with high probability, while non-null hypotheses cannot be confirmed with high probability (.....). In 2002, a group of psychologists launched a new journal dedicated to experimental studies in psychology (.....)
The Journal of Articles in Support of the Null Hypothesis (JASNH) was founded to address a scientific publishing bias (.....). According to the editors, "other journals
and reviewers have exhibited a bias against articles that did not reject the null hypothesis. We plan to change that by offering
an outlet for experiments that do not reach the traditional significance levels (p < 0.05). Thus, reducing the file drawer problem, and reducing the bias in psychological literature. (......) We collect these
articles and provide them to the scientific community free of cost."
H. Johnson, 1999. The Insignificance of Statistical Significance Testing. - Journal of Wildlife Management 63(3):763-772.
"American Psychological Association seriously debated a ban on presenting results
of such tests in the Association's scientific journals. That proposal was rejected, not because it lacked merit, but due to
its appearance of censorship (Meehl 1997). (......)
" The issue was highlighted at the 1998 annual conference of The Wildlife Society,
in Buffalo, New York, where the Biometrics Working Group sponsored a half-day symposium on Evaluating the Role of Hypothesis
Testing (......) Speakers at that session who addressed statistical hypothesis testing were virtually unanimous in their opinion
that the tool was overused, misused, and often inappropriate.
(......) Several interpretations of
P often are made. Sometimes P is viewed as the probability that
the results obtained were due to chance. Small values are taken to indicate that the results were not just a happenstance.
(......) Other times, 1-P is considered the reliability of the result
(......) Alternatively, P can be treated as the probability that the null hypothesis is true. (......) These 3 interpretations are what Carver (1978) termed fantasies about statistical significance. None of
them is true" (.....)
Ipak, taj autor upada u drugaciju zabludu:
"Ordinary confidence intervals provide more information than do
P-values. (......) A confidence interval provides both an estimate of the effect size and a measure of its uncertainty"
Da se ne bi reklo da se "pravim vazan", sutio sam vise od deset godina. Ali, ne radi se o meni, nego o tome, da bi znanost zaista morala biti DOVOLJNO provjerena,
a danas, na zalost, NIJE. Objavljuje se vise radi objavljivanja,
a manje radi znanstvene istine! Zato bi bilo potrebno da svi istrazivaci - bez obzira mogu li slijediti jednostavne
matematicke izvode ili ne mogu - upoznaju
barem konacne zakljucke, te da ih PRIMJENJUJU kod statisticke provjere svojih rezultata!