When analyzing the monthly incidence rate (count of episodes/person-years) using interrupted time series via negative binomial regression, there are some months wherein we don't have an episode of the disease. How should we tackle that, should we impute 0 for the count of an episode for that respective month and some constant value for person-years so that the incidence rate becomes 0 (0/constant). Also, how to treat outliers in the monthly incidence rate data analyzed using the above approach: Interrupted Time Series using Negative Binomial Regression.
... View more