4 ms·

> weren't statistically significant

• Bromskloss 10y ago
  No, give me the numbers first and foremost, and let me draw my own conclusions.
  • PeterisP 10y ago
    That's the whole point - the proper conclusion is "the numbers are meaningless, the experiments we did cannot provide any opinion about the existence, magnitude and direction of such differences", and any conclusions you might draw from seeing some specific numbers would make you less informed than you are right now.
    • Bartweiss 10y ago
      That's not true at all. "Less than significant" is not "the numbers are meaningless", it's "the arbitrary threshold p=.05 was not met". p=.049 is not magic, p=.051 is not worthless. They have far more in common than either does with p=.011, and claiming that giving out specific numbers would make things worse is absurd. I agree that it's bad practice to go "hey, p=.09, but it could be meaningful!" But it's at least as bad to say "we cannot provide any opinion about... such differences". That's simply false, you can provide data and confidence about the topic.
    • lmm 10y ago
      An experiment not finding a statistically significant difference can mean something very different depending on what the power of the experiment was.
    • marcosdumay 10y ago
      To what extent were the numbers meaningless? And how big was the unproven but probably effect so somebody else can try to verify it? Or even not try, discarding it as irrelevant because it was too small? And the most important question that every statistical study should answer: What effect magnitudes were discarded by this research? Claiming insignificant results and refusing to share them is telling your audience "well, we learned something, but we'll hide our findings from you because you are just too stupid to even imagine the kind of thing we learned here." Yes, for some audiences that would be true, but it's something a researcher shouldn't even imagine telling somebody. It's basically incompatible with the kind of thoughts required from researchers.
  • marvin 10y ago
    Not statistically significant means that it is exceedinly likely that the difference was simply due to randomness in the subject selection, and that no difference in fact exists. At this point you're just reviewing the researchers' methodology.
    • Bromskloss 10y ago
      > At this point you're just reviewing the researchers' methodology. That is absolutely part of it.