 |
|
| Science Forum Index » Space - Consult Forum » Paired Comparisons - Tests... |
|
Page 1 of 1 |
|
| Author |
Message |
| Ray Koopman... |
 Posted: Sun Sep 27, 2009 9:31 pm |
|
|
|
Guest
|
On Sep 27, 4:54 pm, Rich Ulrich <rich.ulr... at (no spam) comcast.net> wrote:
[quote:d22a8c03c8][...]
Also -- The result seems to be a 7x15 matrix of 0/1,
for same/different. That's not much information,
especially when it is only 7 subjects.
Seven subjects -- Is that enough to generalize from?
Do the 15 comparisons represent the top of a 6x6 matrix
of objects? - that seems to be needed for the sort of
ranking that is mentioned.
[/quote:d22a8c03c8]
I think he means each subject makes 15_Choose_2 = 105 comparisons.
Each subject would give a 15x15 skew-symmetric matrix whose elements
are +1, -1, or 0. The number of subjects is not stated. |
|
|
| Back to top |
|
|
|
| Rich Ulrich... |
| Posted: Mon Sep 28, 2009 2:53 pm |
|
|
|
Guest
|
On Mon, 28 Sep 2009 00:31:25 -0700 (PDT), Ray Koopman <koopman at (no spam) sfu.ca>
wrote:
[quote:3a46680734]On Sep 27, 4:54 pm, Rich Ulrich <rich.ulr... at (no spam) comcast.net> wrote:
[...]
Also -- The result seems to be a 7x15 matrix of 0/1,
for same/different. That's not much information,
especially when it is only 7 subjects.
Seven subjects -- Is that enough to generalize from?
Do the 15 comparisons represent the top of a 6x6 matrix
of objects? - that seems to be needed for the sort of
ranking that is mentioned.
I think he means each subject makes 15_Choose_2 = 105 comparisons.
Each subject would give a 15x15 skew-symmetric matrix whose elements
are +1, -1, or 0. The number of subjects is not stated.
[/quote:3a46680734]
Ah. Okay, that sounds very reasonable.
Since the OP is searching for the dimension of an "attribute",
it does sound very reasonable that he may be able to get
an answer of "more" or "less" for the comparison on *most*
of the occasions where it is "different."
That leads to a simplification of sorting, since it is likely that
the ends can be determined.
The requested information on sample size or "power" will
depend, very thoroughly, on how distinct the elements
are. That is, if N= 2 subjects agree 100% on all the rankings,
that's a pretty simple ranking, which only leaves the question
of whether there are potential "Subject" differences that were
not invoked.
This sounds somewhat like a problem of Osgood's "Semantic
Differential".
--
Rich Ulrich |
|
|
| Back to top |
|
|
|
|
|
All times are GMT - 5 Hours
The time now is Thu Dec 03, 2009 1:03 pm
|
|