On Tue, 24 Oct 2000 08:37:10 -0400 "John W. Kulig"
<[EMAIL PROTECTED]> wrote:
>
>
> "Kenneth M. Steele" wrote:
>
> >
> > > 2. When using a Likert scale with adjective-modified anchors (as in
> > > "extremely pleasant" or "extremely bitter") - are we permitted to
> > > compare results between subjects/groups? or are we limited to
> > > within-subject comparisons?
> > >
> >
> > This is a much trickier question as the use of the adjectives
> > has caused people to be less likely to use the most extreme
> > values. This would interfere with subjects assigning numbers in
> > an interval-like fashion. This could have several effects
> > depending upon the number of steps on your scale. Should we
> > assume that this individual used these anchors with a 5-point
> > scale?
>
> This weekend I saw the practice of putting the "extreme" adjective not
> at the end of the line - but about 3/4 of the way across - with the end of
> the line hanging into space. This practice - so they say - makes it easier
> to select "extremely."
Consider the following questions...
I. The contents of CUP A taste:
+-------+-------+-------+------+-------+-------+
extremely extremely
bitter pleasant
--VS--
II. The contents of CUP A taste:
+-------+-------+-------+------+-------+-------+
extremely extremely
bitter pleasant
It would still seem to me that version II provides more steps
between the extremes and this is where most responses would lie.
More steps means that there is more opportunity to capture
systematic differences, whether you are doing within or
between-S comparisons and whether you treat the measures as
ordinal or interval. SO it may be true that you are more likely
to see the choice of an "extremely" item in version I, but you
still have lost resolution among the middle items.
Can these practitioners cite empirical work on this scaling
issue?
And what is beyond "extremely"????
Ken
----------------------
Kenneth M. Steele [EMAIL PROTECTED]
Dept. of Psychology
Appalachian State University
Boone, NC 28608
USA
