Conversation

Notices

1. Embed this notice
   Paul Cantrell (inthehands@hachyderm.io)'s status on Tuesday, 01-Oct-2024 04:05:03 JST Paul Cantrell

   • Donald Ball
   • Raven667

   @raven667 @donaldball
   P-values are a curse: it’s easy to produce them for any given piece of data, and •extremely• hard to make them actually mean anything. They tend to just be obfuscation for bad methodology.

   • Embed this notice
     Raven667 (raven667@hachyderm.io)'s status on Tuesday, 01-Oct-2024 04:40:30 JST Raven667

     in reply to

     • Donald Ball

     @inthehands @donaldball sure, people game P-values and the methodology of many scientific papers is not reproducible, but most tech/software metrics don't even reach that level of bullshit.

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Tuesday, 01-Oct-2024 04:46:27 JST Paul Cantrell

     in reply to

     • Donald Ball
     • Raven667

     @raven667 @donaldball
     Several of my statistician colleagues would like us all to more or less ditch p-values for both those reasons: they can be gamed, and they can be applied with so much thoughtlessness that even describing it as “gaming” is giving it too much credit.