Re: Are graded clinical signs more reliable than dichotomized?

On 29 Jun 2006 00:39:19 -0700, "Roland" <rolandersson@xxxxxxxxx>
wrote:

John,

We are talking about a different situation. We are talking about the
description of the intensity of a clinical sign. But also in your
example I think it is true that more grades are better than two? For
colonic cancer the Dukes classification with four grades has a meaning
to describe prognosis and it is currently evolving towards more grades.


You agreed that from a statistical point of view more grades are better
than two. What I want is a reference to support such a statement.

There is a large difference between two and 100 grades. Where is the
optimum? Two? Four? Six?
[snip]

I think that the research you are looking for was done in
the 1930s, for various sorts of scales. Look in OLD textbooks
of psychometrics, if there is not enough in the new ones.

In the 1930s is when Likert showed that using integers works
practically as well as devising more precise weights, etc.; that
is still true unless there are *many* items, and a sizable
norming sample. For Likert-type items, which have symmetric
ends, there is also the question of whether you want to force
a choice away from the center, by not-using an odd number
of choices.

The consensus for steps for psychological ratings of agreement
is 4-7. These have been provided with "anchor" terms, instead
of asking people to "rate from 1 to 7" -- Perhaps, people today
are familiar enough with scales that we can ask with fewer anchors?
- One of the problems, even with 5, is finding terms for
the anchors that are not mis-interpreted by sub-population.
(The word "quite" does not mean entirely the same in British-English
as in American English.)

Elsewhere, you were mentioning "soreness", etc.

It is my impression that there is an enormous literature on
the problems of rating "pain". Some people don't feel much,
because the mind's feedback works in miraculous ways; and some
people stoically don't like to report much. Then there are the
qualitative differences between "itching", "piercing", or whatever.
Try to find the better part of the literature, but I can't help with
that.

--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.