Re: The Mad, Mad, Mad, Mad World or FOUR Afonsos Documented in THIS post.
- From: "Old Mac User" <chendrixstats@xxxxxxxx>
- Date: 22 Aug 2006 06:35:07 -0700
You wrote:
"Why don't you ALL go back to the AOA Asylum of Afonsos and
atay there for awhile so that we can talk about STATISTICS..."
Yes, let's do that.
RF... I've been on vacation and returned last night. It seems that
every time I go away there is another uproar. That conclusion is drawn
from limited data, and I'll not get more data until late October. But
in the meantime let's see if we can slow down to a gentle simmer and
review what's happening here.
I enjoyed your mini-tutorial on hypothesis testing. It was, of course,
accurate and well-stated. So I'll not elaborate further on that.
It seems to me that the current uproar stems from another case of
malcommunication.
As you said in your mini-tutorial, it all begins with a null hypothesis
of the form H0:m1 = m2 (I'm using means here for convenience,
recognizing there are other such comparisons).
But after proper analysis of the data we are responsible for stating
"what this means." In my shop statistics is about communicating
information from one person to another. So, if the test statistic and
its corresponding p-value imply "no detectable difference" then I'm
going to express this outcome in a form such as "we do not have
sufficient evidence to claim there is a difference in (population
means, or whatever applies)." In spite of your elevated blood
pressure, I side with LA at this point. Yes, he could have made his
case with less agitation. Nonetheless, such an outcome... "no
detectable difference"... cannot be used to claim there is no
difference whatsover (in pop. means, or whatever). Equally, such an
outcome must not be distorted to imply that "we have proved m1 = m2."
One common way of settling the matter is to report the confidence
interval on the difference of the means, albeit the words that go with
such a confidence interval must be chosen with care lest they confuse
rather than enlighten.
It's not unusual to find that someone has been careless with words and
has left such a "no detectable difference" outcome framed as if they
they proved there is no difference between m1 and m2. That's no only
misleading, it is simply wrong. More than once I've served as an expert
witness and have relished the opportunity to explain this delicate
matter in the presence of a judge and jury. "Words, like parts, should
fit with precision."
"Statistics" is a lot more than just equations, relationships, numbers,
software, and logic. There comes a point at which we must explain what
all of this means. It's one thing to explain this to fellow
professionals. It's quite another to explain it to those who do not
deal with these matters daily.
So I write this one off as another case of malcommunciation. Your
minitutorial is accurate and even fun to read. It communicates that
part of the story accurately. LA has tried to make the point "but a
failure to establish a difference between m1 and m2 does not imply
there is no difference whatsoever"... and he is correct. These two
pieces work together... they are not opposing pieces of the puzzle.
Be of good cheer... OMU
.
- Follow-Ups:
- Hypothesis Testing. See the FULL lecture in http://tinyurl.com/mlzg6
- From: Reef Fish
- Re: The Mad, Mad, Mad, Mad World or FOUR Afonsos Documented in THIS post.
- From: Kevin E. Thorpe
- Hypothesis Testing. See the FULL lecture in http://tinyurl.com/mlzg6
- References:
- Prev by Date: Re: How could they foresee?
- Next by Date: Re: The Mad, Mad, Mad, Mad World or FOUR Afonsos Documented in THIS post.
- Previous by thread: Re: The Mad, Mad, Mad, Mad World or FOUR Afonsos Documented in THIS post.
- Next by thread: Re: The Mad, Mad, Mad, Mad World or FOUR Afonsos Documented in THIS post.
- Index(es):
Relevant Pages
|
