Mar 22, 2019 20:00:00

Over 800 scientists say the opposition is 'statistically significant'

By

bialasiewicz

The truth should always be one, but even if you experiment to confirm the truth, you will not always get one result. For example, when it is judged whether or not the effect of a drug is not accidental, it is called 'significant' when it is judged statistically that 'the effect of the drug is not accidental' even if it is tested whether it is effective or not. However, 800 scientists have expressed dissenting opinion as being swayed by the word 'significant'.

Scientists rise up against statistical significance
https://www.nature.com/articles/d41586-019-00857-9

'Statistical Significance' Is Overused And Often Misleading: Shots-Health News: NPR
https://www.npr.org/sections/health-shots/20019/03/20/705191851/statisticians-call-to-arms-reject-significance-and-embrace-uncertainty

The founder of statistics at the beginning of the twentieth century, Ronald Fischer developed a technique called significance test . The significant difference test finds the probability that 'the effect of the drug is not effective' even if it is not clear whether the effect of the drug is 'yes' or 'not', and concludes whether 'there is' or 'not'. was. The significant difference test calculates a random variable called ' P value ' from the measurement results of the experiment. For example, if the probability that the experimental result can occur is 95% or more, the P value will be 0.05 or less. By convention, scientists have judged that 'if the P value is less than 0.05, that is, the probability that this event can occur is 95% or more, this experimental result is not accidental' and is 'significant.' Initially, it should have only been shown that 'whether it is significant or not' means 'this experimental result can occur with a probability of 95% or more', but 'whether it is significant or not' gradually affects the conclusion of the research results. It is said that it will be controlled to 'whether research will be announced' and 'whether experiments can be subsidized'.

By monkeybusiness

Mr. Ron Wasserstein, Secretary General of the American Institute of Statistics, 'Don't release the experimental results because researchers who bring the P value closer to their desired values by altering the experimental results, or because they are not significant even if the experiment has significance. There are also researchers. ' Like ordinary people, scientists tend to believe that 'the result is true if it is statistically significant.' According to Blake McSheen of Northwestern University, “It is possible that the experimental results may change depending on the situation of the experiment. It is not a two-choice choice of“ true ”or“ not true ”, but more vague conclusions, such as If you eat food that may be harmful to your health, it is important to think about 'how much the risk to your health is', not 'harmless' or 'harmless'.