> Recently I was having an e-mail discussion regarding the midpoint, as often
> used in repertory grids. The midpoint can denote either a middle value or
> that the construct is not applicable (NA). These are different issues,
> though at times the midpoint is used as a catchall for both issues.
>
> As I understand it options include:
>
> a) forced choice/don't include a midpoint
>
> b) the midpoint is labelled, as a bit like both etc or a bit like neither or NA
>
> c) transform the scale into a unipolar scale, with 0 = NA etc
>
> d) treat NA as a missing value, though I'm not sure of grid programs that
> allow for this.
>
> I would be interested to hear other suggestions or preferences,
Interesting topic. I try to use elements and constructs so that the
grid is working in the range of convenience all the time. Of course,
this isn't always possible. You can get round a lot of the "both
ends apply" by using two constructs with implicit opposite poles
rather than one construct pinned down at both ends. For example,
I try never to use "feminine/masculine" but to use two constructs:
feminine and masculine either with implicit opposite poles or as
"feminine/not feminine". Of course there are potential problems of
"flailing opposite poles" with that but I do think it's well worth trying
to get when you can.
If I got a "both apply" I suppose I might split the element (e.g. E5)
into two, keep the ratings the same for both E5a and E5b for all
other constructs but give then the opposite polar ratings for this
particular construct and look at where that puts them (E5a and
E5b) in the PCA map say.
I tend not to use odd numbers of response levels as it introduces
some of these ambiguities but that's not a magic solution as the
issues are real and can't just be hidden.
I have played with doing what I call a "range of convenience grid"
rather like your option c where the ratings were 1 if the element lay
within the range of convenience of the construct and 0 otherwise.
Then you can get a map of "convenience/applicability" and think
about the relationships and the dimensionality it shows.
Certainly you can use "missing rating" markers and can still do
various analyses, I can see ways in which cluster analysis could
still be applied happily to such a matrix though I don't know of
software that would do it effortlessly (nor should any software do it
effortlessly as we'd stop thinking!) Not using macs (now if you
want a _REAL_ operating system forget this mac/PC rubbish and
use a version of linux .... sorry folks, probably not for grids) I don't
know how REPGRIDII handles this and I'm seriously puzzled about
how it does its PCA for odd or missing ratings.
Brian/Mildred: would you spare a moment to explain that to us? I
for one would really appreciate it.
Very best all,
Chris
PSYCTC: Psychotherapy, Psychology, Psychiatry, Counselling
and Therapeutic Communities; practice, research,
teaching and consultancy.
Chris Evans & Jo-anne Carlyle
http://psyctc.org/ Email: chris@psyctc.org
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%