Goodbye P value: is it time to let go of one of science’s most fundamental measures? [View all]

[n2doc]

How should scientists interpret their data? Emerging from their labs after days, weeks, months, even years spent measuring and recording, how do researchers draw conclusions about the results of their experiments? Statistical methods are widely used but our recent research in Nature Methods reveals that one of the classic science statistics, the P value, may not be as reliable as we like to think.

Scientists like numbers, because they can be compared with other numbers. And often these comparisons are made with statistical analyses, to formalise the process. The broad idea behind all statistical analyses is that they allow the researcher to make somewhat objective assessments of the results of their experiments.

Which drug is more effective?

Scientists often conduct experiments to investigate whether there is a difference between two conditions: do people get better more quickly after taking the blue pill (condition one) or the red pill (condition two)? The most common method for assessing if the pills differ in their effectiveness is to undertake statistical analysis of where some patients were given the blue pill and some the red, and from this determine whether there is strong evidence that one colour is more effective than the other.

To assess experimental results, scientists very often use a “P value” (P is for probability). This is used to show how convincing these results are: if the P value is small, they think that the findings are real and not just a fluke. In our pill example, if P is small this is considered good evidence that there is a difference in effectiveness of the two colours of pill.

more

https://theconversation.com/goodbye-p-value-is-it-time-to-let-go-of-one-of-sciences-most-fundamental-measures-38057