Dear All
This is an intellectual curiosity query.
I have a researcher fitting a first order inverse polynomial to data.
Please do not ask whether this is sensible as this is what their
subject requires them to do. I have grave cautions over doing this
on the size of data set they have but this is not part of the stats.
After much trial and tribulation it became clear that linearising the
formula using 1/x and 1/y was actually statistically sensible i.e. the
errors then tended to be normal.
This is of course non-standard approach statistically and it goes
against up-to-date research advice in the subject.
My instinct is that this is telling me something about the process
going on in the Y measure but I can not think what.
Jean
I am used to often finding that if a log transform normalizes a
variable in a situation then there is some approximation to counting
or growth going on.
------------------------------------------------------
Jean M. Russell M.A. M.Sc. [log in to unmask]
Corporate Information & Computing Services,
University of Sheffield
285 Glossop Road
Sheffield
S10 2HB
United Kingdom
Phone: 0114-222-3098
Fax : 0114-222-3040
|