02-28-2018, 06:13 PM
For those who have the interest and time, a new Nova episode has just ended on PBS this evening - "Prediction by the Numbers". It's all about probability and quite relevant to the discussion here. The video is available online here:
http://www.pbs.org/video/prediction-by-t...rs-hg2znc/
I want to add something to the program's description about "p-values". The narrator said a p-value of 0.05 was statistically significant and means 95% of the time the result of a prediction or experiment is true. Some fields use that number (in particular, medicine), but it's misleading. In my own field, we use a p-value of 0.0000001, in other words, if you repeat the experiment over 3 million times, you'll likely get a significantly different result just once.
http://www.pbs.org/video/prediction-by-t...rs-hg2znc/
I want to add something to the program's description about "p-values". The narrator said a p-value of 0.05 was statistically significant and means 95% of the time the result of a prediction or experiment is true. Some fields use that number (in particular, medicine), but it's misleading. In my own field, we use a p-value of 0.0000001, in other words, if you repeat the experiment over 3 million times, you'll likely get a significantly different result just once.