Dear All,
I have data were repeated measures (84 days) have been taken on each
individual. Only the outcome variable was measured over time, all other
covariates are time-independent.
The outcome variable is binary (symptom or no symptom present) and I
have calculated a summary measure per person of representing the
proportion 'p' of days symptom free, p=r/n per person, where r are the
no. of days where no symptom was present and n represents the overall
no. of days in the study.
Before I progress with more a complex model using the raw data, I wish
to use my summary measure 'p' in a logistic regression model. I would
like to know if this can be carried out, as my outcome is not the
typical p=proportion of subjects with 'no symptoms'. If I could carry
out this regression would the interpretation be slightly different?
I have a second summary measure of the binary outcome, the 'median
response' per person. Would this more suitable?
Thank you
Susan White
|