Discrete PD Outcome Population analysis requires a good understanding of important statistical concepts. In addition, discrete data are more difficult to visualize using standard exploratory analysis. The session will focus on Multi-Level Categorical Responses where data are levels (i.e., 0,1, 2) with a model being probabilistic, Count data where data are integers (0,1,2…) but the basic rules of addition/multiplication apply and time to events, where the data is itself the event. Multi-level categorical responses can be modeled using either proportional odd models with or without discrete Markov processes, continuous Markov based processes, or finally for multiple linked categorical responses, using item response theory. Count and time to events require the use of hazard rate concept, integrated hazard rate and survival distribution.
All these concepts are complex and therefore I will use intuitive statistics rather than formal statistical theory to give access to a better understanding by a broader audience. Finally, case studies will be presented all along the session.
Learning Objectives:
Understand the different types of categorical responses we encounter in Pharmacometrioc and how to differentiate betweem then
Optimally use special type of exploratory tools to better select PD iniital parameter estimates
Optimally develop identifiable models that involve categorical responses
Optimal present goodness of fit plots for categorical responses
