At 8:24 +0000 3/9/99, Catherine Kay wrote:
>I am working on a project that collects a lot of data through
>questionnaires. many questions are scored on a 5 point scale from strongly
>agree to strongly disagree, the mid point being 'uncertain' or 'don't
>know'.Objections from medical and statistical collegues that 'don't know' is
>not a midpoint, from a statistical point of view, are I feel,logical. I feel
>that without this option, respondents may be forced into expressing an
>opinion they do not hold, or may not answer. Apart from accepted problems
>with this type of measurement anyway, can anyone clear up the conceptual
>problem of giving a score to 'uncertain'.Oppenheim acknowledges that neutral
>is not necessarily the midpoint, yet if it is to be included, how can one
>score it?
>Many thanks for any help here.
>From Catherine Kay
I think you'll get objections from a number of areas (psychology included)
on the nature of the scale and treating don't know as a mid point. The problem
is that the debate between those who will require a forced choice and omit
the midpoint and those who prefer to give respondants an 'opt out' midpoint
will continue to rage.
Equally there would be concerns raised regarding the traditional scoring of
1 through 5, raised from a discussion of whether or not the data is truely
interval in nature. The safest way to analyse the data is to treat it as
ordinal
and work from that starting point (that said, there are a large number of,
often highly respected people, who will happily score 1-5 and be done with it).
On the crux of the mail 'the score to uncertain' it could be argued that this
isn't only a position of uncertainly but also one that reveals neutral feelings
towards the item being assessed. If this is the case, then it would appear
appropiate to score the response. The problem is that you can't establish
whether the response is 'neutral' or 'don't know/uncertain' so either approach
will lead to some false scoring.
It might be worth loking at some of the work by Steve Newstead on the labels
applied to scales and the differences that these have on respondents patterns
of answers, it may touch further of the conceptual problem that you are
interested in.
good luck
Andy
=======================================================================
Andy Morley
Department of Psychology, University of Plymouth,
Drake Circus Plymouth, Devon, PL4 8AA
Phone: (00 44) (0) 1752 233179 -- Fax: 01752 233176
http://psy.plym.ac.uk/staff/amorley/amhp.html
=======================================================================
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|