11-20-2015 05:27 AM - edited 11-20-2015 06:21 AM
I am working on age specific seroprevalence data, trying to derive past incidence. I use the proc logistic to fit a binomial model with a log-log link function. The procedure is presented in SAS user guide for the logistic procedure. In my result, and in the SAS example, the first age group when considering one year per age group has a higher infection rates than the others (almost double, when all the other group are quite close).
Could it be associated with the fact that the age-group representing the time exposure is longer for the one year old group? The hypothesis is that the one year old group has an incidence which is calculated for an exposure of 2 years (from 0 to 1 year and 23 months) when for the other group the incidence is calculated with a time exposure of one year, i.e., the time between each group ( 1 to 2 years old, 3 to 4 years old, ...)
Has anyone did this kind of analysis? Is the hypothesis for the differences in the first age group valid? If yes, any idea of how to correct that?