Dear colleagues,
I am currently using a Generalized Linear Model to evaluate
associations of morbidity with farm management factors in cattle
for a relatively small data set. The morbidity rate is approximately
10%, and I have been assuming that the error distribution follows a
binomial distribution and using a logit link function.
I am finding that the residuals are skewed with the majority of
negative residual values close to zero, and positive residual values
substantially greater than zero. What I need to know is much can I
trust the odds ratios and tests of statistical significance from the
analysis.
Any advice would be appreciated? My email address is
[log in to unmask]
Best wishes,
L. Malcolm Hall, Ph.D.,
Epidemiology Unit,
Scottish Agricultural College,
Stratherrick Road,
Inverness, IV2 4JZ
Tel: 01463 243030
Fax: 01463 711103
e-mail: [log in to unmask]
|