Jointly modeling transition rates of multi-state cognitive outcomes with random
  
   Lead Investigator:    Elizabeth Schofield
   Institution      :    Byrd Alzheimer's Institute
   E-Mail           :    eschofie@health.usf.edu
   Proposal ID      :    88



   Proposal Description:
   Multi-state modeling is a technique used to simultaneously model transition rates between diagnostic states. This method is prone to high variability in estimates due to the number of possible transitions and the need for separate parameters for each covariate-transition combination. Random effects models adjust for heterogeneity in effects, but require a near-continuous random variable in order to estimate such effects. In this project we will use a surrogate variable and a multi-state variable to jointly model transition rates and the random effects which affect transition rates. Specifically, we will jointly estimate transition rates between normal, MCI, and dementia states with random effects shared between these transition rates and those reflected in decline of other surrogate variables such as MMSE and trail-making tests.