Re: Odd p values in t test
- From: Richard Ulrich <Rich.Ulrich@xxxxxxxxxxx>
- Date: Tue, 27 Sep 2005 17:27:19 -0400
On 27 Sep 2005 13:29:26 -0700, "Rerics" <mccres@xxxxxxx> wrote:
[snip, paired t-test details]
> >
>
> I guess I just thought that the Var2-Var1 of 1.1, 1.5 and 1.1 (which
> represents only a ~2% change) with such a small sample size would give
> a much higher p value than 0.01.
>
Keep in mind that the "t-test" does not know that the
numbers are only 2% change; it just sees consistency.
A t-test would be exactly the same if those numbers
were (11, 15, 11) or (110, 150, 110).
If *you* know, as a scientist, that the changes are
small compared to the "noise" in the measurement,
or the noise in the system,
then you have reason to reject the outcome of the test.
Did that just *happen* to be a very small standard-error
of the change?
--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.
- Follow-Ups:
- Re: Odd p values in t test
- From: LWn
- Re: Odd p values in t test
- References:
- Odd p values in t test
- From: Rerics
- Re: Odd p values in t test
- From: LWn
- Re: Odd p values in t test
- From: Rerics
- Odd p values in t test
- Prev by Date: Re: Odd p values in t test
- Next by Date: Re: analysis of variance
- Previous by thread: Re: Odd p values in t test
- Next by thread: Re: Odd p values in t test
- Index(es):
Relevant Pages
|
|
