Interview with Dr. Henry Bauer - Part 1

147 Replies, 19999 Views

(2017-10-27, 09:19 PM)Laird Wrote: Fair enough, Chris. I think that's a justifiable view, although I'm willing to extend Linda the benefit of the doubt and grant that she's being honest when she says she did mean that power decreased even though she only mentioned effect size decreasing (because that would be the case if the number of trials remained unchanged, which is what she claims her statement assumed).

The trouble is that would imply she thought it would be appropriate to test a weak effect by using the same number of trials as for a strong effect. She knows better than that.
Linda, you identified the statement in question, in your post #101: "smaller effects (decreasing power) increase the likelihood that that positive results are false-positives, even (especially) in the setting of very low p-values".

This is clearly the statement to which I was referring when I wrote 'We simply interpreted your statement differently: you didn't explicitly refer to power, only to "smaller effects"'.

Your response, "This is untrue", is, then, itself untrue! Clearly, you didn't explicitly refer to power in that statement - because the only explicit reference (the parenthesised comment) is one that you have added; it wasn't present in the original on which Chris and I originally based our interpretation.

So, to attempt to justify your response, "This is untrue", you then switch to a different statement in a different thread, which makes a different point.

Linda, I'm not going to continue this - it would go on forever, and I don't have the patience for it.
(This post was last modified: 2017-10-27, 10:12 PM by Laird.)
[-] The following 2 users Like Laird's post:
  • tim, Doug
(2017-10-27, 10:06 PM)Chris Wrote: The trouble is that would imply she thought it would be appropriate to test a weak effect by using the same number of trials as for a strong effect. She knows better than that.

Fair point again.
It is not an "error" to refer to effect sizes with respect to false positive discovery rates. "Power" is a value which is calculated from the characteristics of a study, and one of those characteristics is "effect size". You can't change the effect size and not also see a change in the power. You can arrange the characteristics of a study so that the calculation returns a specific value. But then you would need to refer to each of the characteristics individually which allowed you to get there. You certainly wouldn't assume, after hearing that one of those characteristics had been changed, that the other characteristics had been changed in some sort of unspecified manner, as well.

I specifically offered a link which referred to power, including references to power under small, medium and large effects. And I specifically referred to power, and how it depended upon (among other things) effect size.

And Laird and Chris finally seem to be on the same page with respect to this - they recognize that my statement which makes this explicit for them is true. So why make a fuss about this?

Linda
(This post was last modified: 2017-10-29, 02:25 PM by fls. Edit Reason: clarify terminology )
I just dislike people being dishonest.
This post has been deleted.
(2017-10-27, 10:24 PM)Chris Wrote: I just dislike people being dishonest.

Me too.

What exactly are you trying to claim I lied about? That I failed to tell everybody that effect sizes don't have an effect on false-positive discovery rates, because it should be assumed that sample sizes and alpha levels can be and have been changed to accommodate any effect size? And that we should always assume that this is the case, regardless of what that hack Jacob Cohen finds otherwise?

Linda
(This post was last modified: 2017-10-29, 02:27 PM by fls. Edit Reason: clarify terminology )
(2017-10-27, 10:35 PM)fls Wrote: What exactly are you trying to claim I lied about? That I failed to tell everybody that effect sizes don't have an effect on false-positives, because it should be assumed that sample sizes and alpha levels can be and have been changed to accommodate any effect size? And that we should always assume that this is the case, regardless of what that hack Jacob Cohen finds otherwise?

There's very little point in explaining yet again why what you said was wrong. I think more than enough has been said about this for anyone who is (or was at some point) interested to get the point.
This post has been deleted.
(2017-10-27, 11:17 PM)fls Wrote: Ah I see. There wasn't really anything I was wrong about.

Can anyone remain under the delusion that it's possible to have a sensible discussion with this person?
[-] The following 1 user Likes Guest's post:
  • tim

  • View a Printable Version
Forum Jump:


Users browsing this thread: 1 Guest(s)