"Kenneth M. Steele" wrote:
>
> I agree, but the major problem here is caused by comparing
> across nonequivalent groups.
>
> I see this problem ignored most often when people are talking
> about "gender differences." Consider the following
> stem-and-leaf plots of answers to the question---
>
> Is "Doom" a gory computer game?
>
> 1 2 3 4 5 6 7
> Not Very Medium Very
>
> Men Women
>
> 7 7 +++
> 6 +++ 6 +++++++
> 5 +++++++ 5 ++++++++++
> 4 ++++++++++ 4 ++++++
> 3 ++++++ 3 ++
> 2 ++ 2 +
> 1 + 1
>
> Notice that the distribution of Men's answers are shifted
> towards "Not" by 1 item but are otherwise identical to the
> Women's answers.
>
> It doesn't matter whether you treat the data as interval or
> ordinal (the usual argument involving Likert-type scales). The
> summary statistic would indicate that men rate the game as less
> gory.
>
> As John indicated, the problem with this conclusion is that we
> don't know whether males and females use the scales in the same
> manner. The difference may only mean that men and women have a
> systematically different definition of the words "not very."
This is exactly the problem as I underdstand it. I am in the middle of
tracking down references on this problem. I heard last weekend that either_SS
Stevens or Likert claimed Likert (LICK-ert) scales were for within-subjects only.
If a group is homogeneous, thee should not be a problem - but we don't always know
if a group is homogeneous.
These issues came up in Linda Bartoshuk's talk on tasters and non-tasters at
NEPA. She used an example similiar to yours. NT = nontasters, people with few
taste buds, T(asters) have a medium amount, and ST(supertasters) have alot of
taste buds. One of her graphs showed hypothetical results of these 3 groups of Ss
tasting a bitter flavor. This is the graph of REALITY:
ABSOLUTE
MEASURE OF
SENSATION: >--------LOW-------->----------HIGH--------->
NT *********** "Very Strong"
T *********************"Very Strong"
ST *****************************"Very Strong"
"Very Strong" would represent the upper limit of each group to experience taste
sensations. When the data is compressed onto the same scale, differences between
the 3 groups will be decreased. A group difference could disappear entirely - or
(in an extreme case) reverse direction.
--
---------------------------------------------------------------
John W. Kulig [EMAIL PROTECTED]
Department of Psychology http://oz.plymouth.edu/~kulig
Plymouth State College tel: (603) 535-2468
Plymouth NH USA 03264 fax: (603) 535-2412
---------------------------------------------------------------
"What a man often sees he does not wonder at, although he knows
not why it happens; if something occurs which he has not seen before,
he thinks it is a marvel" - Cicero.
