Fabio Manganiello on Nostr: To me this paper states the obvious, but sometimes it's better to state the obvious ...
To me this paper states the obvious, but sometimes it's better to state the obvious to make sure that everybody in the scientific community is on the same page:
1. p-points are a useful statistical tool to infer significance, but they aren't the ONLY statistical tool.
2. Setting the threshold for statistical significance at p=0.05, with everything above it bring inconclusive/insignificant and everything below being surely statistical significant, is a completely arbitrary convention that does more harm than good. Probabilities are continuous, and any attempt to discretize them (or, worse, build dichotomies on arbitrary thresholds) ends up distorting results, discarding potentially useful results, or enforcing unethical practices ("just massage the data until you get that p=0.05").
3. It's not the job of researchers to draw conclusions from their researches. All the studies should be published, in a way or another, and conclusions should only be drawn after quantitatively comparing results from different papers.
https://phys.org/news/2023-12-abandon-null-hypothesis-significance-default.htmlPublished at
2023-12-14 11:53:10Event JSON
{
"id": "bf83f740fc934dd0f0844d10fa505d4e0aff540bec0ee9ec010fdab51ee509d6",
"pubkey": "678fbdf04a787406ea6ccc5fd35c1cf57ac74ea9d0aa81df88f7a941f57e75e3",
"created_at": 1702554790,
"kind": 1,
"tags": [
[
"proxy",
"https://social.platypush.tech/users/blacklight/statuses/111578630773947728",
"activitypub"
]
],
"content": "To me this paper states the obvious, but sometimes it's better to state the obvious to make sure that everybody in the scientific community is on the same page:\n\n1. p-points are a useful statistical tool to infer significance, but they aren't the ONLY statistical tool.\n\n2. Setting the threshold for statistical significance at p=0.05, with everything above it bring inconclusive/insignificant and everything below being surely statistical significant, is a completely arbitrary convention that does more harm than good. Probabilities are continuous, and any attempt to discretize them (or, worse, build dichotomies on arbitrary thresholds) ends up distorting results, discarding potentially useful results, or enforcing unethical practices (\"just massage the data until you get that p=0.05\").\n\n3. It's not the job of researchers to draw conclusions from their researches. All the studies should be published, in a way or another, and conclusions should only be drawn after quantitatively comparing results from different papers.\n\nhttps://phys.org/news/2023-12-abandon-null-hypothesis-significance-default.html",
"sig": "a7ade96ca5c4c1b1e7a92fa7ed3588f4b6593304539e6940c435dd036f99d82affe70a88b0330449c518b69205dcb21ac5369bf77c65823645c30fbc2f1a89c9"
}