Dear All
BACKGROUND
A study with two treatment groups, with pre and post treatment biopsies.
Primary outcome was mast cell count.
The cell count data was standardised to the area within which they were
observed, ie units = number of cells per 0.1 square mm.
I do not have the original raw data of counts and corresponding area over
which they were observed. The actual areas observed could be less than or
greater than 0.1 square mm.
I intend to perform a Poisson regression on the area-standardised cell
counts setting the offset to 1, hence ln(1)=zero.
QUESTION
I would like to estimate the possible bias in parameter estimates
(specifically the treatment parameter) and in the standard errors, when I
use the area-standardised data vs using the raw data.
Thank you for your anticipated interested and help.
Seasons Greetings
Tri
Tri Tat
Medical Statistician
________________
Research Services
Statistical Support
Chelsea Wing, RBH
Tel: 020 7351 8748
Fax: 020 7351 8578
|