|
Gabriel S. Jacobs@gsjphd.bsky.social |
Simple isn't necessarily inferior! Not only are you demanding a bunch of additional tests beyond the scope of the research question, but you want it to be a kitchen sink regression. If you throw enough terms into that analysis, eventually you'll find ONE that correlates by random chance. p-hacking!
1 replies 0 reposts 1 likes
|
|
Gabriel S. Jacobs
@gsjphd.bsky.social
[ View ] |
Rather than throw every possible orbital variable in until something pings p<0.05, here's the responsible way to do things: A. Identify a specific variable (or combination) with a mechanistic explanation for why it would have an effect & test it, or B. Add a penalty to models based on # of variables
1 replies 0 reposts 0 likes