…
Innen medisin og andre samfunnsfag bruker man ofte statistisk testing av hypoteser…
Man klarlegger "den signifikante sannhet" ved hjelp av den magiske p-verdi som skal være mindre enn 0.05.
Adam Kucharski skriver i boka : Proof - the Uncertain Science of Certainty, i kapitlet Big Lies:
"One of the biggest risks to any scientific study is randomness. Indeed, much of modern statistics was developed to avoid coincidence for truth. In turn, wider science adopted one statistical concept in particular—the p-value—as the main defence against coincidence. The smaller the p-value, the more confident we can be that the result isn't down to chance. Based on Ronald Fisher's (1890-1962) popular p-value cut-off of 5 per cent, it might be tempting tothink that at least 95 per cent of research findings report a genuine effect. The reality, though, is very different..
The popularity of Fischer' p-value cut-off means that researchers often struggle to publish findings that come in above this influental number. Most scientific journals want to publish new discoveries, not inconclusive results. Some researchers have responded by getting creative in their descriptions. Studies failing to meet the traditional cut-off have employed euphemisms ranging from 'essentially significant' (p-value of 10 per cent) and 'arguably significant' (p=9 per cent) to ' approximating significance' (p= 7 per cent)."
Det eneste man kan oppnå med statistisk hypotesetesting er å påvise at null-hypotesen sannsybligvis er feil, dvs. en falsifiseing av null-hypotesen. Man påviser aldri at noen hypotese er riktig, uansett hvor mange hypoteser man tester…‽
…
P-values simplified, WMed,
Ray Scott Percival: Confirmation versus Falsificationism, The Encyclopedia of Clinical Psychology, 2015-01-23 (PDF)
What Most Doctors Don’t Understand About False Positives, Living Systems, ????-??-??
…